Monetary Policy-Making under Uncertainty

Conference organised jointly by the

European Central Bank (ECB)

and the Center for Financial Studies

of the University of Frankfurt

on 3 December 1999

in Frankfurt am Main

Impressum

Monetary Policy-Making under Uncertainty

EditorsEuropean Central Bank,Center for Financial Studies

PublishersIgnazio AngeloniFrank SmetsProf. Dr. Axel Weber

Typeset & EditingRoberta Ruben

DesignStählingdesign, Darmstadt

PrintWerbedruck PetzoldJune 2000

December 3-4, 1999

ProgrammeFriday, December 3, 1999

Editors Introduction

Introduction by W. Duisenberg (European Central Bank)

Policy Panel

Introduction by J.B.Taylor, (Stanford University), Chair

Otmar Issing (European Central Bank)

Steven Cecchetti (Ohio State University)

Charles Freedman (Bank of Canada)

Leo Leiderman (Bank of Israel)

Lucas Papademos (Bank of Greece)

Dinner Speech

Alan Blinder (Princeton University)

Session Summaries

L.E.O. Svensson (Stockholm University and Princeton University) and M.Woodford (Princeton University): „Indicator Variables for Optimal Policy“

T. Sargent (Stanford University and Hoover Institution) andIn-Koo Cho (University of Illinois, Urbana Champaign): „Escaping Nash Inflation“

A. Levine,V.Wieland and J.Williams (Federal Reserve Board): „The Performance of Forecast-Based Monetary Policy Rules under Model Uncertainty”

A. Orphanides (Federal Reserve Board): „The Quest for Prosperity without Inflation“

Saturday, December 4, 1999

G. Rudebusch (Federal Reserve Bank of San Francisco): „Assessing Nominal Income Rules for Monetary Policy with Model and Data Uncertainty“

Shorter Presentations

M. Ellison and N.Valla (European University Institute):„Learning, Uncertaintyand Central Bank Activism in an Economy with Strategic Interactions“

P. Geraats (University of California, Berkeley):„Transparency and Reputation:Should the ECB publish its Inflation Forecast?“

B.Winkler (European Central Bank):

„On the Need for Clarity in Monetary Policy-Making“

P. Schellekens (Harvard University, London School of Economics and UFSIA):

„Caution and Conservatism in the Making of Monetary Policy“

U. Söderström (Sveriges Riksbank):

„Monetary Policy with Uncertain Parameters“

6

12

18

20

30

38

48

56

64

76

80

83

86

89

92

93

94

95

96

(measurement error, forecast error, para-meter uncertainty, model uncertainty) onhow central banks should respond to deve-lopments in the state of the economy wereanalysed in the more scientific-oriented ses-sions.

The first academic presentation was apaper on “Indicator Variables for OptimalPolicy” by Lars Svensson (StockholmUniversity) and Michael Woodford

(Princeton University), and it was discussedby José Viñals (Banco de España) andGuido Tabellini (Bocconi University).Svensson and Woodford derive the optimalweights on monetary policy indicators inmodels with forward-looking variables andonly partial information about the state ofthe economy. To demonstrate the pro-perties of their model, they discuss anexample of optimal monetary policy withunobservable potential output and a partial-ly observed cost-push shock.

In the second session, Thomas

Sargent (Stanford University and HooverInstitution) and Ju-Koo Cho (University ofIllinois, Urbana-Champaign) presentedtheir paper on “Escaping Nash Inflation”,which was discussed by Ramon Marimon

(European University Institute) and James

Stock (Harvard University). The Sargentand Cho model assumes that the monetaryauthority controls the inflation rate up to arandom disturbance, but does not know thetrue data-generating mechanism. Instead ituses a good fitting approximating learningmodel.The authors show that in such a fra-mework least squares learning ensures con-vergence to rational expectations equilibri-um. But, under fixed-gain recursive learningschemes that discount past observations, theso-called “escape dynamics” can drive theeconomy away from a rational expectationsequilibrium. The authors argue that suchdynamics could explain the great inflationexperience of the 1970s.

The third paper on “The Performanceof Forward-looking Monetary Policy Rulesunder Model Uncertainty” was delivered byAndrew Levin, Volker Wieland andJohn Williams (Federal Reserve Board)and was discussed by Charles Bean

(London School of Economics) and Stefan

Gerlach (Bank for International Settle-ments). Levin, Wieland and Williams com-pare the performance of outcome- andforecast-based rules in four differentmacro-econometric models of the U.S.

7

Editors Introduction

As a new central bank operating in acompletely new environment, the EuropeanCentral Bank (ECB) still faces greater-than-usual uncertainty about the state of the euro-area economy and the working of the mo-netary policy transmission mechanism. Inaddition, the ECB must establish a trackrecord for maintaining price stability, andthe gradual acquisition of that credibilitycan be an additional source of uncertainty.

Both forms of uncertainty have relevantimplications for the design and frameworkof monetary policy and need to be betterunderstood by both decision-makers andobservers.For this reason,the European CentralBank and the Center for Financial Studiesorganised a conference on “Monetary Policy-Making under Uncertainty”, which washeld in Frankfurt on 3 and 4 December1999, and featured contributions from high-level policy makers and top internationalacademic experts in this field.This bookletcontains the speeches delivered at the con-ference, the statements of the policy panel-lists, and a summary of the main paperspresented and their discussions.The papersin their entirety can be downloaded from

the ECB’s webpage at http://www.ecb.int(Working Paper Series Section) as well asCFS’ web-site at http://www.ifk-cfs.de.

The conference, which attracted morethan 160 participants, was opened on Friday3rd December with a welcoming address bythe ECB President Wim Duisenberg. Hestressed that this first, large, open conferen-ce, co-organised by the ECB, is yet anothersignal of the ECB’s commitment to a conti-nuous and active exchange with the acade-mic world.

The morning session, which was chai-red by John Taylor (Stanford University),started with a panel of high level policy-makers. The experts on the panel wereSteven Cecchetti (Ohio State Universi-ty), Charles Freedman (Bank of Canada),Otmar Issing (ECB), Leonardo Leider-

man (Bank of Israel), and Lucas Papa-

demos (Bank of Greece).The panel discus-sed the implications of different forms ofuncertainty for the design and frameworkof monetary policy from the specific angleprovided by their own policy experience.

The qualitative and quantitativeimpact of various forms of uncertainty

6

On each of them, Blinder gives his viewsbased on his combined academic andpolicy-making perspective.

The paper by Glenn Rudebusch

(Federal Reserve Bank of San Francisco) on“Assessing Nominal Income Rules forMonetary Policy with Model and DataUncertainty” was discussed by Ben Mc-

Callum (Carnegie Mellon University) andHenrik Jensen (University of Copen-hagen). Rudebusch analyses two issues con-cerning nominal income targeting rules formonetary policy. First, he examines theperformance of such rules over a range ofplausible empirical models – especiallymodels with and without rational inflationexpectations. Second, he analyses the per-formance of these rules, in real time, usingthe type of data that is actually available topolicy-makers rather than final reviseddata. The paper compares optimal mone-tary policy rules in the presence of suchmodel uncertainty and real-time datauncertainty and finds only a limited role fornominal output growth.

The conference concluded with aseries of five shorter presentations, whichwere jointly discussed by Alex Cukierman

(Tel Aviv University), Marvin Goodfriend

(Federal Reserve Bank of Richmond), andCarl Walsh (University of California atSanta Cruz).

The paper by Martin Ellison andNatacha Valla (European UniversityInstitute) on “Learning, Uncertainty andCentral Bank Activism in an Economy withStrategic Interactions” presents a stylisedmodel in which learning, uncertainty andstrategic behaviour play a role.The model isestimated with G7 data and used to exami-ne the optimal degree of activism of a cen-tral bank. The authors show that a centralbank which takes into account that itsactions could affect learning, may choose tobe less active than a central bank that igno-res learning effects.

The paper by Ulf Söderström

(Sveriges Riksbank) on “Monetary Policywith Uncertain Parameters” presents a sim-ple dynamic macroeconomic model andshows that uncertainty about structuralparameters does not necessarily lead to amore cautious monetary policy. In particu-lar, when there is uncertainty about thepersistence of inflation, it is optimal for thecentral bank to respond more aggressively

9

economy. They also investigate how robustthese rules are with respect to modeluncertainty.The paper shows that in all fourmodels forecast-based rules yield at bestonly small benefits in stabilising inflation,output, and interest rates relative to opti-mised outcome-based rules that respond toinflation, the output gap, and the laggedinterest rate. This is even true in the twolarge-scale models which contain literallyhundreds of state variables and allow forsignificant lags until the maximum effect ofa policy change on the economy is felt.Moreover, forecast-based rules may inducemultiple equilibria with self-fulfilling ex-pectations, if the forecast horizon is suffi-ciently long.

The paper by Athanasios Orphanides

(Federal Reserve Board) on “The Quest forProsperity without Inflation”, which Jordi

Galí (Pompeu Fabra and New York Univer-sity) and Paul de Grauwe (Leuven Uni-versity) commented on, also dealt withactivist monetary policy rules in which thecentral bank responds to inflation and thelevel of economic activity. Orphanides re-constructs real-time data that was availableto U.S. policymakers from 1965 to 1993.Using an estimated model, he then per-

forms counterfactual simulations underalternative informational assumptions re-garding the knowledge policymakers canreasonably have had about the state of theeconomy when policy decisions were made.When realistic informational assumptionsare used, findings favoring activist policiesare overturned in favour of prudent policiesthat ignore short-run stabilization concernsaltogether. The evidence points to misper-ceptions of the economy's productive capa-city as the primary underlying cause of the1970s inflation and suggests that apparentdifferences in the framework governingmonetary policy decisions during the 1970s,compared to the more recent past, havebeen greatly exaggerated.

A highlight of the conference was thedinner speech given by Alan Blinder

(Princeton University). After noting thatthe notion of monetary policy under uncer-tainty is somewhat redundant (“was thereever such a thing as monetary policy-making under certainty?”), Blinder listed 15issues that he considers central to moderncentral banking. The issues can be groupedinto three categories: 1) institutional design;2) tactics for operating in the markets; 3)the model of the transmission mechanism.

8

The conference organisersIgnazio Angeloni (ECB),Frank Smets (ECB) and Axel A.Weber (CFS and Goethe-University)

11

to shocks than under certainty equivalence,since the central bank this way reducesuncertainty about the future developmentof inflation. Uncertainty about other para-meters, in contrast, acts to dampen thepolicy response.

In his contribution “Caution and Con-servatism in the Making of Monetary Policy”Philip Schellekens (London School ofEconomics) asks whether or not caution andconservatism improve the conduct of mone-tary policy-making. In his model he exami-nes the implications of caution for the cre-dibility and flexibility of monetary policyand suggests a reconsideration of the rolefor conservatism as an element of delegati-on.The key finding of the paper is that the con-servative-central-banker approach neednot imply sub-optimal output stabilisation.

The paper by Petra Geraats (Uni-versity of California at Berkeley) on “Trans-parency and Reputation: Should the ECBPublish its Inflation Forecast?” aims at ex-plaining why central banks may want toabandon secrecy in monetary policy andbecome very transparent by focussing onthe disclosure of central bank forecasts.Themodel shows that transparency leads to

lower inflation and gives the central bankgreater flexibility to respond to shocks in theeconomy. Furthermore, transparency makesit easier for a central bank to build reputa-tion and in order to achieve such benefitsfrom transparency it is shown to be neces-sary to publish the conditional central bankforecasts for both inflation and output.

Finally, the paper by Bernhard

Winkler (ECB) entitled “On the Need forClarity in Monetary Policy Making” statesthat greater transparency about the policy-making process is one way of minimisingthe potential for misunderstanding betweenthe ESCB and financial markets and thepublic. He proposes distinguishing betweendifferent aspects of transparency, and em-phasises the particular importance of “clari-ty” in communicating central bank policy.“Clarity” means that information needs tobe simplified, packaged, processed andinterpreted in order to be understood. Inthis context, it is pointed out that the ECBfaces an extraordinary communication chal-lenge due to it being both a new and asupranational institution which must makeitself understood vis-à-vis a fragmented setof audiences in a multi-cultural environ-ment.

10

Introductory Statement

12

Willem F. Duisenberg,President of the

European Central Bank

place, accountable to the public at large, notto the limited circle of academic econo-mists. I should imagine that the answer isobvious to most of you. Let me, neverthe-less, expand somewhat on the most impor-tant reason, and this relates to what AlanBlinder, who will be the speaker at tonight’sdinner, called the large, and sometimesunexploited, potential gains from trade inideas between practitioners and academics.

Central banks are among the mostintensive users of economic research. It istherefore of utmost importance that theECB should continuously update its know-ledge of, and expertise in, the most recenttheoretical and empirical research findings.We should never be, as Keynes put it, un-witting “slaves of some defunct econo-mist”, while pretending to be exempt fromintellectual influences. Just as most firmsneed to update regularly their computerand software systems in order to remaincompetitive, a central bank needs to ensu-re that it understands and can implementrecent economic research in order toachieve its primary goal of price stability, inthe most efficient way. Of course, just asmost firms would not abandon their oldtime-honoured systems for new, untestedones without gathering evidence of theirbenefits and reliability, central banks needto be very cautious about jumping on thebandwagon of any new paradigm that aca-demic researchers or other observers mayprovide.

The usefulness of economic researchin monetary policy-making is very clear at

all levels, from the institutional design ofcentral banks to the details of day-to-daymonitoring.We do not need reminding thatit was the revolution in rational expectati-ons and the results concerning the timeinconsistency of optimal policies, togetherwith the experience of the great inflationof the 1960s and 1970s and its associatedcosts, that contributed to a wave of changesin central bank charters based on the prin-ciple of independence and a clear focus onprice stability. These principles also repre-sent the foundations of the Treaty on Euro-pean Union, which established the Euro-pean System of Central Banks.With this inmind, it was, for example, decided rightfrom the outset that the Eurosystem shouldclearly define its primary objective of pricestability as an increase in the HarmonisedIndex of Consumer Prices of below 2%.

A similar mixture of sound theoreticaland empirical research, and practical experi-ence underpins the Eurosystem’s monetarypolicy strategy, which, as you are aware, isbased on a prominent role for money (in the form of a reference value for the growthof M3 of 41⁄2 %) and a broadly based asses-sment of the outlook for price develop-ments.The money indicator provides a refe-rence point for the evaluation of the pros-pects for price stability. Its usefulness isbased on the long-term relationship bet-ween money and prices, which characteri-ses virtually all economic models and whichhas been extensively illustrated in empiricalstudies. This relationship reflects the factthat inflation is ultimately a monetary phe-nomenon.

Ladies and gentlemen,

It is a great pleasure to welcome youhere to Frankfurt on the occasion of theconference entitled “Monetary policy-mak-ing under uncertainty”, organised jointly bythe European Central Bank (ECB) and theCenter for Financial Studies of the Univer-sity of Frankfurt.

As you can imagine, I have had theopportunity to live through many “firsts” inrecent times. It all started on 9 June 1998,when the Governing Council of the ECBmet for the first time. Then, on 4 Januarythis year, the first transactions in euro wereconducted in the money and foreign ex-change markets and the first payments ineuro were settled in the new TARGETsystem.Three months later, the first decisi-on to change official interest rates wasadopted by the Governing Council on 8April. And I could continue. In the light ofthe unique historical journey upon whichwe embarked as a consequence of the esta-blishment of a single currency in 11 sover-eign countries, the smooth and successfulstart of Stage Three is a tribute to the dedi-cation and relentless efforts of many of mycolleagues at the ECB, at its precursor, theEuropean Monetary Institute, and at thenational central banks, and of many othersinvolved outside the central banking com-munity.

This event is another important “first”that is worth celebrating: the first, large,open conference organised by the ECB. Ishould like to use this occasion to thank

Professor Weber from the Center forFinancial Studies for co-organising thisevent with us. This conference – and I amsure there will be many more to follow – isyet another signal of our commitment to acontinuous and active exchange betweenthe ECB and the academic world. As manyof you know, such conferences are not theonly way in which we interact with acade-mics. Many of you have already visited theECB to present your views and discuss issu-es of policy interest with our economists.This year alone we have hosted around 50presentations in the context of our InvitedSpeaker and Lunchtime Workshop pro-grammes. In spite of the heavy workloadfaced by the ECB staff, such presentationsare generally well attended and very muchappreciated.

In addition, ECB economists presenttheir work at academic conferences on aregular basis. The ECB has also launched aWorking Paper Series via which our rese-arch is disseminated and to which the spea-kers at this conference will also contribute.Finally, this summer we started a newGraduate Research Programme in whichstudents are expected to pursue part oftheir dissertation, work here at our head-quarters, on topics that are relevant fromthe viewpoint of ECB policies. One suchstudent, Petra Geraats, will tomorrow pre-sent her work at this conference.

Why do we invest so much time andresources in this exchange with the acade-mic world? After all, as an independent pan-European institution, the ECB is, in the first

1514

touch upon the other leg of this exchange.How can central bankers contribute to theacademic world? Again the answer seemsobvious. By bringing together researchersand central bankers in conferences like this,by inviting academics to give presentationsat our central bank and engage in discussi-ons with our economists, by bringing aca-demics from their ivory towers into thehustle and bustle of real-world policy-making, we hope that we can inspire themto go back to their universities and researchinstitutes and work on the issues that con-front us in our daily policy-making process.Some of these issues are very practical, likedevising new econometric techniques todetect structural change. Others are empi-rical, like understanding movements involatile asset prices, or examining the impli-cations of much-needed structural reformsin labour and goods markets. Finally, someof the most important issues for centralbanks are analytical: for example, under-standing the role of money and credit in thetransmission mechanism of monetary policy, or the issues arising in conductingmonetary policy in conditions of low infla-

tion. All these issues show that whoeverthought monetary policy would becomeboring once price stability was achieved,was clearly way off the mark.

After all, one of the few things we eco-nomists know with certainty, is that weknow little without uncertainty. It is, there-fore, entirely appropriate that the first aca-demic conference organised by the ECBshould deal with the topic of “Monetarypolicy-making under uncertainty”. I wishyou very interesting and stimulating discus-sions and I look forward to witnessing theresults.

17

While the reference value for mone-tary growth is a useful and robust policyguide in many ways, and ensures that theECB maintains a medium-term perspec-tive, it is clear that the ECB, like any othercentral bank, needs to take into account allthe information it has at its disposal. Nocentral bank can afford to rely on a singleindicator or a single rule.The second pillarof our stability-oriented monetary policystrategy provides a framework in which thatinformation can be processed and presen-ted. The most important guiding principleis that information should be used to theextent that it is relevant for the assessmentof future price developments and the risksto price stability.

And this brings me to a third area inwhich economic research has an importantrole to play: the monitoring of indicatorsand of their implications for the outlook forprice developments. Let me focus onmodel-based analysis. Clearly, such analysis,including model-based forecasts, can be auseful tool for interpreting the availableinformation in a consistent and coherent

framework. However, these forecasts areonly as good as the underlying models thatare used. Moreover, different models maybe used to answer different questions. We,at the ECB, are committed to developingand maintaining a set of tools that are usefulfor analysing the euroarea economy, andexamining the implications for future infla-tion. This is, however, not a trivial taskgiven the large uncertainties that we arefacing due to the establishment of a multi-country monetary union. Not only can weexpect some of the historical relationshipsto change due to this shift in regime, butalso, in many cases, there is a lack of com-parable historical and cross-country dataseries that can be used to estimate suchrelationships. We are availing ourselves ofsome of you as consultants in this difficultprocess. However, the bottom line is that,in spite of all the progress made, it takestime to develop such tools, and that some ofthese uncertainties will never disappear.

I have mainly spoken to you about theimportant input that economic researchprovides in central banking. Let me briefly

16

1918

Opening Remarks

Thank you very much for an excellentbeginning to this conference on PolicyMaking Under Uncertainty. I want to thankthe European Central Bank and the Centerfor Financial Studies for organizing a reallyvery interesting conference – and for invit-ing me to it.

The first session of the conference is apanel of policy makers discussing the sub-ject of Monetary Policy Under Uncertainty.It is an honor for me to chair this openingpanel with such a distinguished group ofpolicy makers.

I think it is an excellent idea to start aconference on research on policy uncer-tainty by hearing from the policy makerswho actually have to make decisions underuncertainty. There is nothing like having tomake a decision about whether to raise orlower an interest rate – an action that affec-ts many people’s lives – to focus one’s mindon the importance of uncertainty. And so,

we will begin by hearing about what it islike to make decisions under uncertainty.

Another reason that it is an excellentidea to start the conference by hearing frompolicy makers is that the ultimate goal ofthe policy research to be presented later isto help inform policy makers and to helpthem make better policy decisions. So, inmy view, this beginning puts the emphasison the right priorities.

We have a panel of five experiencedpolicy makers: Prof. Otmar Issing from theEuropean Central Bank, Prof. StevenCecchetti who has just stepped down asResearch Director at the Federal ReserveBank in New York and is now at Ohio State,Dr. Charles Freedman of the Bank ofCanada, Dr. Leo Liederman from the Bankof Israel, and Governor Lucas Papademosfrom the Bank of Greece.And we are going to begin with ProfessorIssing.

John B.Taylor,Stanford University

2120

Otmar Issing,Member of the Executive Board of the

European Central Bank

The monetary policy of the ECB in a world of uncertainty

the same time, so close to the frontier ofacademic research and so relevant for ourdirect concerns. Our desire was, and still istoday and tomorrow, to give an opportuni-ty to central bankers to learn from recentacademic research on this issue and, viceversa, to enable academic economists tobenefit from the awareness of the practicalconcerns of central bankers.

2.The many aspects of uncertainty inmonetary policy-making

Academic economists hardly need anyreminding that central bankers have tomake decisions in a world of pervasiveuncertainty. However, while the academicprofession has made tremendous progressin analysing risk in well-defined stochasticeconomies, the “Knightian” uncertainty thatconfronts central bankers is altogether ofanother dimension. Among the variousforms of uncertainty that central bankersface, the uncertainty about how the policyinstrument affects inflation and economicactivity – the monetary transmissionmechanism – and the uncertainty about thecurrent state of the economy – the data –appear to weigh particularly heavily.

Central bankers need to have a goodunderstanding of the timing and the ultima-te effects of changes in the monetary policyinstruments on inflation and economic ac-tivity in order to be successful in conduc-ting monetary policy. For this purpose,monetary policy-making requires morethan just the qualitative information thattheory provides.They must have quantitati-ve information about magnitudes and lags,

even if that information is imperfect. How-ever, within the profession of central ban-kers – probably even within a particularinstitution – there is no common view onthe appropriate specification of a model sui-table for the analysis of monetary policyissues.

Given the high degree of model uncer-tainty, central bankers highly welcome therecent academic research on the robustnessof monetary policy rules across a suite ofdifferent models. Some of that research willbe presented at this conference.

Even if there were a consensus on anappropriate model, there would remainconsiderable uncertainty about the model’sparameters. Since Brainard (1967) it is wellknown, that this form of uncertainty provi-des a rationale for a prudent, gradualist ap-proach to monetary policy-making.The po-licy relevance of this result has nicely beenreiterated by the former Vice-Chairman ofthe Board of Governors of the Federal Re-serve System Alan Blinder (1999) who con-cludes that “a little stodginess at the centralbank is entirely appropriate.”

However, research, which will be pre-sented tomorrow, shows that this resultdepends on the exact source of parameteruncertainty. In particular, if the uncertaintyconcerns the persistence of inflationarydevelopments, central banks may be welladvised to be firmer in their policy respon-se. Similarly, adopting robust control theo-ry Sargent (1999) finds that parameteruncertainty does not necessarily imply a

23

1. IntroductionAs a central banker directly involved in

monetary policy-making, I have been dea-ling with uncertainty and its consequencesfor a large part of my professional life.From my experience as a member of theBoard of the Bundesbank, I have vividmemories of the challenges posed byGerman reunification and the turbulencesurrounding the ERM crises. But neverhave I felt the impact of uncertainty so acu-tely as in the weeks that preceded and fol-lowed the introduction of the euro and thebirth of the single monetary policy. Now,after almost a year, we feel more comforta-

ble: we have learned a lot from that experi-ence. But exactly one year ago this week,when the Governing Council made its finalannouncements on the monetary policystrategy and was getting ready for theChangeover Weekend, the uncertainty wasat its peak. Nothing could be taken forgranted, no matter how careful had the pre-paratory work been.

It was precisely in those weeks, and forthese reasons that the idea of hosting an aca-demic conference on “Monetary policy-making under uncertainty” came to ourminds. No other subject seemed to us, at

22

Given this uncertain environment,however, the ECB did not have the optionto wait for even some of the uncertaintiesto resolve. Instead, the ECB designed itsmonetary policy strategy to guide mone-tary policy in the best possible way throughall the deep waters that inevitably surroundthe transition to the single currency.

Dealing with these deep waters had atleast three important implications forformulating the strategy. Let me turn brief-ly to them.

3. CredibilityFirst, in an environment of uncertain-

ty it is important to establish and maintainthe credibility of the central bank for achie-ving price stability. High credibility willreduce the probability that unexpecteddeviations in inflation from price stabilitywill be interpreted as a change in objectivesrather than the result of shifts in the under-lying relationships.Although, in the end, it isthe central bank’s actions and performancethat will determine its reputation, the man-date of the Treaty – to maintain price stabi-lity as the primary objective – and its insti-tutional set-up provide a solid foundationfor building such a reputation for the newlyestablished ECB.

Moreover, the pre-commitment ex-pressed through the announcement of thestability-oriented strategy helps to preservethe anti-inflationary reputation from itsprecursors, the national central banks. Forthis, and the previous reason, it was alsoimportant that the ECB clearly defined

what it means by price stability. Someobservers have criticised this definition fornot being precise enough. The definition,however, contains two very precise state-ments.The ECB does not consider inflationabove 2% as price stability; and the ECBdoes not consider deflation as price stabili-ty. Being more precise, in the form of a pointtarget, is likely to be counterproductive,exactly because of some of the uncertaintieswe will discuss at this conference, such asmeasurement bias in the price index.

A measure of the credibility the ECBenjoys in the markets can be extracted fromthe yields on long-term bonds. If one takesa look far enough ahead, beyond the horiz-on of business cycles, changes in long-termnominal interest rates typically reflect mar-kets’ perceptions of long-term inflation risks.If the central bank is credible, long termrates will not move very far away fromlevels consistent with maintained prospectsof price stability; they will quickly jump tohigher levels if credibility is lost. If I take alook at long-term bonds denominated ineuro, I can conclude that the ECB has alrea-dy earned a considerable level of credibilitygiven the particularly high degree of uncer-tainty it faces.

Despite the increase over the course ofthis year, which seems to reflect mainly glo-bal factors, long-term interest rates in theeuro area appear fully consistent with a pro-longed period of price stability. In fact, themost recent decrease in connection withthe Governing Council’s interest rate deci-sion on 4 November confirms this view.

reduction in the responsiveness of mone-tary policy.

Beyond the need of a suitable model, athorough assessment of the current state ofthe economy is central to the formulationof monetary policy. Among the multitudeof useful indicators the concept of the out-put gap and, intimately related, that ofpotential output, plays an important role inmeasuring future inflationary pressures.Unfortunately, potential output is unobser-ved and, as with estimates of other unob-served variables such asthe NAIRU or the equi-librium real interestrate, there is no obviousway in which it can bedefined and estimated.

Moreover, sincecentral banks need tooperate in real time, thedata used in the assessment of the currentstate of the economy are incomplete andsometimes subject to substantial revision.For the U.S., Orphanides and van Norden(1999) have conducted a comparative studyof alternative methods for estimating theoutput gap in real time. Research presentedlater today shows that taking into accountthe sizeable measurement error derivedfrom these real-time estimates leads to asignificant deterioration of feasible policyoutcomes and causes efficient policies to beless activist.

All in all, academic research as well asthe practice of central banking seem to reaf-

firm the famous dictum by Milton Fried-man (1968) questioning activist monetarypolicy. This dictum seems true, even morefor the ECB’s monetary policy, since thelevel of uncertainty which the ECB faces onaccount of the transition to Stage Three ofEMU is even higher than the level of uncer-tainty which central banks face in normaltimes. The ECB is confronted with a histo-rically almost unique regime shift ac-companying the introduction of the singlecurrency.

With regard tothe operation of thesingle monetary poli-cy, we can only safelysay that we know, atbest, the broad con-tours of the euro-areatransmission mecha-nism right now. Whatwe can say for sure,

however, is that there is a considerable like-lihood that the way monetary policy istransmitted may change, making the task ofthe ECB even more difficult. For example,the restructuring and the intensification ofthe competition within the banking systemwill deeply affect the interest rate channelgiven the dominant role it plays for financi-al intermediation in the euro area. Theuncertainties surrounding the data aremagnified by a lack of area-wide time series.Therefore, econometric analysis has to makeinferences on the basis of “fictitious”, aggre-gated time series that pre-date the formati-on of EMU, and the way in which these timeseries are constructed is not beyond dispute.

2524

and also a lack of strong empirical evidenceat the time when the decision was taken,explains why the Eurosystem decided not toadopt a monetary targeting framework.

But, the Eurosystem decided alsoagainst adopting an inflation-targeting fra-mework. Instead, by introducing a referen-ce value for the growth of a broad monetaryaggregate – not a target in the traditionalsense –, the two-pillar strategy is designedto ensure that a balance is reached betweenthe information which is provided by thebroadly-based assessment of the prospectsfor price stability, and the medium-terminformation which is contained in themonetary aggregate.

Not surprisingly, it takes some timeuntil it is widely recognised that the stra-tegy chosen by the ECB is neither monetarytargeting nor inflation targeting, nor even amixture of these two approaches wellknown to observers. It is a new strategydesigned for a unique situation with whichthe ECB was confronted before the start ofStage Three.

Nevertheless, I found recent critiqueof not having adopted either a monetary oran inflation targeting strategy not very con-vincing since it was not accompanied by anyevidence of why the reason for the cautioustwo-pillar approach of the ECB were notjustified.

5. CommunicationI have so far emphasised the virtues of

flexibility. But there is a clear limit – or

counterbalance – to it: flexibility must beaccompanied by a continuous effort towardsopenness and clarity in communication.

High uncertainty complicates commu-nication with the public and the financialmarkets. As a result, one cannot rely onsimple communication devices, but needsto tell the full story. Some observers havewrongly taken the comprehensive asses-sment of the prospects for, and risks to,price stability to be synonymous with aninflation forecast, which is customarily atthe centre of direct inflation targeting stra-tegies. However, the broadly-based asses-sment comprises an analysis of such a widerange of indicators and information varia-bles – including various internal and exter-nal forecasts – that it cannot be reasonablysummarised in a single number or chart.

Instead, the Eurosystem’s strategyprovides an honest account of all the piecesof information that have been taken intoaccount in the decision-making process.The ECB’s Monthly Bulletin summarisesthe analysis of the information leading tothe decisions of the Governing Council, andprovides the most recent statistical infor-mation on the euro-area economy. More-over, immediately after the first GoverningCouncil meeting of each month, the ECB’sPresident and Vice-President together holda press conference where they comment onthe meeting, explain the decisions taken,and then answer questions.

The two-pillar strategy of the Euro-system helps organising all the available in-

27

4. FlexibilitySecond, in an environment of high

uncertainty about the economy, flexibility –some would call it discretion – is impor-tant. Monetary policy decisions mustalways remain a valid option which resultsfrom a complex evaluation of empiricalresults, theoretical reasoning, and judgmen-tal inputs. This is why central banking hassometimes been described as an art, and thisis why a central bank cannot afford to bestrictly bound by any simple policy rulewhich may be optimal for a given structureof the economy, but breaks down as soon asthe structure changes.

More generally, no central bank canafford to disregard information.That is, nocentral bank can escape the need to conti-nually assess the state and the operation ofthe economy using a multitude of indicatorsand information variables, and base theirpolicy decisions on the comprehensiveassessment of that information for the risksto price stability. The second pillar of ourstrategy ensures that the ECB takes all thisinformation into account with its relevancefor future inflation as guiding principle. Atthe same time, however, a prudent centralbank should recognise that much of theinformation at its disposal is fraught withconsiderable noise. As we learned fromFriedman, it does not pay to be too activist ina very uncertain world (and various papersin this conference will illustrate this point).

Once we recognise that informationalproblems, among others, limit the scope foractivist policy, a robust policy guide with a

medium-term orientation would be a use-ful tool. In the Eurosystem’s strategy thereference value for the growth of the broadmonetary aggregate M3 plays such a role. Itis based on a simple framework, which isrobust to important forms of data uncer-tainty (though not all of them). It is inten-ded to help analyse and present the infor-mation contained in M3 in a manner thatoffers a coherent and credible guide formonetary policy aimed at price stabilityover the medium term. More basically,monetary growth in line with the referencevalue should be consistent with the mainte-nance of price stability at that horizon.

Of course, the practical usefulness ofmoney growth as a medium-term policyguide will also depend on the existence andthe stability of a long-run relationship bet-ween money and prices as typically embo-died in structural models of long-runmoney demand. Recent empirical studies atthe ECB have provided evidence on the exi-stence and the stability of the latter. Morerecent academic research tends to showthat the real money gap, i.e. the differencebetween current real money holdings andits long-run level, is a good predictor offuture inflation.

We have fairly good reasons to be con-fident that the long-run relationship bet-ween money and prices is not significantlyaffected by the transition to Stage Three ofEMU, i.e., that it is robust as it has proved,ubiquitously, to be in the past. However, wemust also be aware of possible structuralshifts in the future.This form of uncertainty,

26

ReferencesBlinder, A.S. (1999), Central Banking in Theory

and Practice, (MIT Press) Cambridge, MA.

Brainard,W. (1967), “Uncertainty and the Ef-

fectiveness of Policy”,American Economic Review

Papers and Proceedings, 57, 411-425.

Friedman, M. (1968), “The Role of Monetary

Policy”,American Economic Review, 58, 1-17.

Orphanides, A. and S. van Norden (1999), “The

Reliability of Output Gap Estimates in Real

Time”, Finance and Economics Discussion Series,

1999-38, Federal Reserve Board,Washington, DC.

Sargent,T.J. (1999), Comment on L. Ball, in J.B.

Taylor (ed.), Monetary Policy Rules, (University

of Chicago Press) Chicago.

29

formation in a way which permits both tomore efficiently structure the internal deci-sion-making process and to communicatethese decisions to the public.The monetarypolicy strategy is thus an important vehicleto explain policy decisions and thereby toalso affect market expectations and, indi-rectly, economic behaviour and outcomes.

To stabilise market expectations andthereby increase the effectiveness of itsmonetary policy actions, the ECB should donothing to add to the level of uncertaintyconfronting the private sector of the econo-my. More precisely, the ECB needs to actand to convey its actions to the public in sucha way that errors in the market expectationsof what the ECB is going to do are minimised.

6. Conclusion To conclude, in dealing with the pervasi-

ve uncertainties that surround the intro-duction of the single currency, the Euro-system’s strategy is able to combine the sophi-sticated demands to modern central bankswith the traditional prudence that centralbanks need to adhere to in order to avoid beingthemselves a source of monetary instability.

Our strategy is credible and flexible atthe same time and allows for a timely res-ponse to a changing environment whilekeeping the objective of price stability inclear focus. It communicates the commit-ment to price stability by providing a cleardefinition. This helps to anchor inflationexpectations and also preserves the anti-inflationary reputation inherited from itsprecursors, the national central banks.

The two pillars – a reference value forthe growth of a broad money aggregate anda broad based assessment of the outlook forinflation – are used to explain monetarypolicy decisions necessary to maintain pricestability.The prominent role of money – assignalled by the announcement of the refe-rence value – is rooted in robust theoreticaland empirical arguments accumulated overmany decades of research.

Of course, in the uncertain circum-stances surrounding the early years of StageThree, the ECB is particularly aware of thechallenges posed by a changing environ-ment. In this regard, without losing sight ofa few fundamental principles that have beenknown for a long time in the discipline, theECB is looking forward to a fruitful discus-sion about the advances of research on theoptimal design of monetary policy underuncertainty which will be presented at thisconference.

28

30

Stephen G. Cecchetti,Ohio State University

Monetary Policy Making and Uncertainty

31

The directive agreed to on 5 October1999 contained an upward bias, stating that:“In view of the evidence currently available,the Committee believes that prospectivedevelopments are more likely to warrant anincrease than a decrease in the federal fundsrate operating objective during the inter-meeting period.”

While the minutes of this meetingbecame publicly available only on 18November 1999, and so the exact wordingof the directive was only known 6 weeksplus two days later, a brief statement relea-sed the day of the meeting making it clearthat, this language had been adopted.Following a brief comment about how “adecreasing pool of available workers willingto take jobs” posed risks to the inflation out-look, the statement closed with the follo-wing remarks: “Against this background, theCommittee adopted a directive that was bia-sed toward a possible firming of policygoing forward. Committee membersemphasized that such a directive did not sig-nify a commitment to near-term action.TheCommittee will need to evaluate additionalin-formation on the balance of aggregatesupply and demand and conditions in finan-cial markets.”

Following the most recent meeting on16 November, when the FOMC raised thefederal funds rate target by 1/4 percentagepoints to 5 1/2 percent, the announcementof the interest rate change was accompaniedby a comment that the bias had been retur-ned to neutral.That is, the final sentence ofthe directive in place today reads: “In view ofthe evidence currently available, theCommittee believes that prospective deve-lopments are equally likely to warrant anincrease or a decrease in the federal fundsrate operating objective during the inter-meeting period.”

Does the FOMC’s current announce-ment policy clarify policy maker's intenti-ons, reducing the uncertainties in the envi-ronment? Or, does this scheme actually ob-fuscate the FOMC’s intentions, makingpolicies more opaque and increase risks sur-rounding future policy actions? There arenumerous difficulties with the currentdisclosure procedures. Primary amongthem is that the statements clearly mean dif-ferent things to different people.

A primary reason for this is that theFOMC is actually not clear about its ownobjectives. Unlike the European Central

33

Most uncertainty is unavoidable. Forinstance, we know that the current, extraor-dinary, boom in the U.S. will eventuallycome to end, we just don't know when.Weknow that the sustainable growth rate of theU.S. economy has increased over the pastfive years, we just don't know how much.The list goes on.The job of central bankersis to conduct monetary policy in order topromote price stability, stable growth, and astable financial system. They do this in anenvironment fraught with such unavoidableuncertainties. But in conducting policy, thereis one uncertainty that policy makers canreduce: the uncertainty they themselves crea-te. Everyone agrees that monetary policymakers should do their best to minimize thenoise their actions add to the environment.The essence of good, transparent, policy isthat the economy and the markets respondto the data, not to the policy makers. Doesthe U.S. Federal Reserve's Federal OpenMarket Committee (FOMC) meet this test?Or, do the current FOMC practices addnoise to the economic environment? Theseare the questions faced by the committeechaired by Federal Reserve Board GovernorRoger Ferguson, that is due to recommendpossible changes as early as the next FOMCmeeting on 21 December 1999.

Current practices were put in place on22 December 1998, when the FOMC an-nounced that it would periodically releasestatements explaining some of its policyactions, and clarifying the committee's thin-king about likely future events. Beginningwith the announcement on 18 May 1999,there have now been 5 such statements, onefollowing each of the FOMC meetings bet-ween May and November. To understandthe history of these statements, it is impor-tant to begin with the FOMC’s policydirective. The directive has two parts, thefirst is a brief description of the currentstate of the economy, and the second setsthe target for the federal funds rate and 1Cecchetti is Professor of Economics, OhioState University, and former Director ofResearch at the Federal Reserve Bank ofNew York. This essay is based on a remarksprepared for the policy panel at the jointEuropean Central Bank/Center for FinancialStudies conference on “Monetary Policy-Making under Uncertainty,” Frankfurt,Germany, 3-4 December 1999.

1. then describes the committee’s con-sensus about the likely future course ofinterest rates.This last portion of the direc-tive has become known as the “tilt” or “bias”.

32

Bank, whose primary objective is price sta-bility, defined as inflation in theHarmonized Index of Consumer Prices ofless than two percent, or the Bank ofEngland, where the Monetary PolicyCommittee is charged with maintaininginflation of 2 1/2 percent in the RetailPrice Index, the Federal Reserve has nopublicly stated objectives. The language ofthe Humphrey-Hawkins Act, which is verylong on goals and very short on details, is allthat we have.As a result, not even the mem-bers of the committee seem to agree on thespecific goals they are striving to achieve. Iscurrent policy targeting inflation? If so, atwhat level? Using whatmeasure? Some articu-lation of these goalsseems essential if wecan ever hope for cla-rity and transparencyfrom the FOMC.

A second pro-blem is that the bias inthe directive has no agreed upon interpreta-tion, even within the FOMC.Taken literal-ly, the second sentence of the final para-graph of the directive is a statement aboutthe intermeeting period.That is, about like-ly interest rate actions prior to the next for-mally scheduled FOMC meeting. Withinthis context, some people seem to thinkthat the bias gives discretionary authority toAlan Greenspan, as the Chairman of theFOMC, to change interest rates on his own.This interpretation has it that if the bias isneutral, as it is today, then the Chairmancan move the federal funds rate by 1/4 per-

centage points in either direction. With abias, he would be able to go 1/2 percenta-ge points in the direction of the tilt, but notthe other way. Even if there were an under-standing that this were the case, and there isnot, it is inconceivable that any Chairmanwould use such power without extensiveconsultation with the members of the com-mittee.

On a practical level, the tilt has beenused as a consensus building device. TheFOMC is unlike both the ECB council,whose votes are never published, and theBank of England’s Monetary Policy

Committee, who havea clear willingness todisagree publicly onthe current stance ofpolicy. In contrast, theFOMC both publishesits votes (in minutesreleased two days fol-lowing the next mee-ting), and has a traditi-

on of public consensus. For there to bemore than two dissenting votes, out of thetwelve voting members (when there is a fullcomplement), is nearly unheard of.Dissents, particularly by the Governors orthe President of the Federal Reserve Bankof New York, would be viewed by manyobserves as a revolt against the Chairman.The bias in the directive acts as mechanismwhereby the Chairman insures that there isnot significant dissent.When there is signi-ficant disagreement on the committee, theChairman will offer those with only milddifferences of views a biased directive as a

way of expressing themselves withoutactually voting against his position.

The history of the statement is that,after a newspaper story disclosed the direc-tive soon after the May 1998 meeting, trig-gering a sharp drop in stock prices, a num-ber of the participants in the FOMC pro-cess expressed their dissatisfaction.

While some people approved of thepolicy of disclosing the directive with a sub-stantial lag, when it had become essentiallyirrelevant, others thought it best to anno-unce the bias immediately after it was adop-ted.The current selective release of the biasand the statement is a result of the discussi-ons that ensued during that year.The inten-tion was to reduce the speculation aboutthe committee’s position, and increase thetransparency of policy making.

My view is that the current strategy hasbeen increasing the uncertainty created bythe FOMC, not decreasing it, as had beenhoped. There are several difficulties. First,the bias refers only to the intermeeting peri-od, while the statements that have beenreleased thus far use wording like “policygoing forward,” implying a longer horizon.

Furthermore, by referring to selectiveconcerns in the outlook, such as the recentfocus on labor productivity and the pool ofavailable workers, the statement suggeststhat future actions are predicated largely onthe data that is in a small subset of what willbecome known before the next meeting.Finally, the statement gives the impression

that the committee has been able to sum-marize its likely reaction to future events ina few sentences (the statements have beenbetween 3 and 6 sentences). Surely none ofthese is either accurate or possible.

As for the tilt itself, it is important toask whether it serves any useful purpose atall. In the absence of a more detaileddescription of the sense of the committee,the bias is very difficult to interpret. It isnot, as some market participants seem tobelieve, a conditional vote about the nextmeeting.That is to say, when the committeesets a biased directive, it is not voting tochange interest rates in the future. Thiswould be very bad practice, and no oneshould be advocating it.

I have tried to argue that the currentpolicy of including a bias in the FOMCdirective, with selective public release ofthe bias and an accompanying statement onthe day of the meeting, increases confusion.Where should we go from here? First, I pro-pose the elimination of both the bias and thestatement as they are currently constituted.The statement is too short to properly con-vey the sense of the committee, and theonly purpose the bias serves is internal con-sensus building. If there is a bias, the marketparticipants will speculate about it, and tryto figure out what it is, and so there is littlejustification for not announcing it.

Looking to other central banks, we seeseveral alternatives. The first is the ECB'sapproach, following that of the Bundes-bank, where the President holds a press

3534

37

conference at the conclusion of each coun-cil meeting. This promotes transparencyand enhances understanding. The give andtake of a news conference allows for adescription and justification of the decisionthat has been taken, and its accompanyingsubtleties. I am in favor of the Chairmanholding a news conference following eachFOMC meeting.

My preferred solution, and my recom-mendation to the Ferguson Committee, isthe publication of the minutes of the mee-ting be moved forward. The Bank ofEngland publishes detailed minutes of theMonetary Policy Committee’s deliberationstwo weeks following each meeting. This isan excellent communications device, as itcontains details that make policy decisionsmuch easier to understand, with the con-cluding section of the minutes would con-tain a clear synopsis of the committee’sviews on the future.The two week lag mayseem like an eternity to financial marketparticipants, where two minutes can be along time, but it is the minimum that is phy-sically possible. This short wait is likely tobe worth the overall reduction in the uncer-tainty of monetary policy.

36

3938

Charles Freedman,Deputy Governor, Bank of Canada

Monetary Policy Making and Uncertainty

developments in the United States, which isby far our largest trading partner. Given theexternal forecast, we then assess futuredevelopments in the Canadian economy,most notably in aggregate demand andinflation. The forecast of inflationary pres-sures, in turn, indicates the monetarypolicy actions required to bring inflation tothe target rate or to hold it there. In thiscontext, we are faced with additive, multi-plicative, data, and model uncertainty.

Additive uncertainty comes from theprojections of external and domesticdemand and supply developments, orshocks. In particular, are recent demandshocks likely to be long-lasting or transi-tory? Or, put another way, is the momen-tum in the economy likely to persist or tofade? An example from the recent pastmight be instructive. When the Asian crisisfirst broke out, our interpretation was thatit would have relatively little direct effecton the Canadian economy because ourtrade with that region of the world was sosmall. While we recognized that the Asiancrisis might have indirect effects on Canadathrough its possible impact on the U.S. eco-nomy and on the prices of raw materials,we initially did not expect these factors tobe very important. But as the full magnitu-de of the decline in the Japanese economybecame apparent and as the prices of rawmaterials were increasingly affected by theglobal slowdown, our views concerning theimpact of the developing crisis on thedemand for Canadian goods and servicesand on the future rate of inflation changedconsiderably.

Additive uncertainty is the simplestform of uncertainty, and it typically requi-res the central bank to take action based oncertainty equivalence. In practice, however,interpreting whether an unexpected outco-me is the result of a demand shock, a supp-ly shock, or both; deciding whether theshocks are likely to be transitory or long-lasting; and making clear the degree ofuncertainty associated with the interpreta-tion of such shocks when communicatingwith the public and with markets are noteasy tasks. They are central to the foreca-sting process and take considerable timeand effort on the part of policy-makers.

Consider, for example, a demandshock that has typically been autocorrelatedin the past but for which, in the current cir-cumstances, no information is available re-garding the likely degree of autocorrelati-on. More concretely, suppose that on one-half of the occasions when the shock occursit lasts two quarters, while for the otherhalf it lasts only one quarter. The certainty-equivalent optimal policy is always to act asif the shock will continue into the secondquarter but with a weight of one-half. Onone-half of the occasions, the central bankwill therefore underreact, and inflation willrise above the target, and on one-half of theoccasions, it will overreact and inflationwill fall below the target.

At times in the past, in the face of thiskind of uncertainty, central banks have cho-sen to underreact and to wait and seewhether the shock actually persisted intothe second quarter. In circumstances where

41

The views in this note are the author`s and do

not represent those of the Bank of Canada.

I would like to thank Pierre Duguay, Paul

Jenkins, David Longworth, Tiff Macklem, Jack

Selody, Gabriel Srour, and Gordon Thiessen for

comments on an earlier version of these remarks.

In my remarks, I will focus on howvarious forms of uncertainty have affectedthe practical aspects of monetary policy-making in Canada.The discussion, of cour-se, reflects the large and growing body oftheoretical literature on policy-makingunder uncertainty, since the way policy-makers interpret the reality that they face isstrongly influenced by how that reality ischaracterized in theoretical models. I willalso be emphasizing some of the issues thatface a central bank in a small open economywith flexible exchange rates.This approachcontrasts with most of the literature, whichfocuses on models of closed economies.

I should note, by way of introduction,that the policy-maker’s role is not just toassess the economic situation and then take“appropriate” action. It is also to explainthe reasons for the action, and, indeed, forthe whole policy approach, to the publicand to the markets in a convincing manner.The importance of communication cannotbe overstated. Public support for the broadapproach to policy is essential in a demo-cracy. And the effectiveness of monetarypolicy is appreciably increased if the expec-tations of the markets and of the public arein line with those of the authorities. But, aswe shall see, it is precisely the various forms

of uncertainty – additive uncertainty, mul-tiplicative uncertainty, data uncertainty, andmodel uncertainty – that make it difficult todecide on the need for action and themagnitude of that action, and then to con-vince markets and the public that the actionis appropriate.

Let me begin with a framework, likethe one in place in Canada, in which thecentral bank has a target for the rate ofinflation and responds to deviations of theforecast rate of inflation from that target sixto eight quarters in the future.The focus onforecasts of inflation six to eight quarters inthe future takes into account the lags bet-ween monetary policy actions and theireffects on inflation. This forward-lookingapproach allows policy-makers to move in atimely manner and thus to avoid generatingexcessively sharp movements in output bytheir actions. Of course, using an inflationtarget as the nominal anchor entails opera-ting under a regime of flexible exchangerates.

The inflation-targeting frameworkclearly places a great deal of emphasis onthe central bank’s forecast of inflation andon its model of the transmission mechanismbetween its actions and their eventual effecton inflation. Both of these involve consider-able uncertainty.

Let me begin with the inflation fore-cast. In a small economy like Canada’s, thestarting point for the forecast is an asses-sment of developments in the world eco-nomy, and we pay particular attention to

40

Market uncertainty regarding the typeand duration of shocks affecting the eco-nomy could cause difficulties for the cen-tral bank in its conduct of monetary policy.To lessen these difficulties, the central bankshould be as transparent as possible aboutthe uncertainties surrounding its outlookfor the economy and its assessment of infla-tionary or disinflationary pressures.

How can a central bank communicatethe nature and extent of the uncertainty itfaces to the public and to the markets? Oneway is to formally set out the probabilitydistribution of the forecast path for inflati-on, as is done, for example, in the fan chartin the Bank of England’s Inflation Report.Less formally, the central bank can talkabout a range for output growth and infla-tion, thereby giving the public and the mar-kets some indication of the degree ofuncertainty around the forecast. Moreover,it can describe the kinds of risks surroun-ding the forecast to draw attention to theconditionality of any forecast and to theuncertainty about the assumptions thatunderlie the base case. It can also presentalternative scenarios based on differentpaths for exogenous variables, such asdemand growth in the country’s tradingpartners. That said, there can be a tenden-cy at times on the part of markets to viewforecasts as being more precise than thecentral bank intends.Therefore, no oppor-tunity should be lost to emphasize thenature and extent of the uncertainties sur-rounding forecasts and, hence, policy deci-sions, and the particular sources of thoseuncertainties.

In summary, while dealing with additi-ve uncertainty is simple in theory, in prac-tice the difficulty of interpreting the type ofshocks, drawing the appropriate conclusionfor the required policy action, and explai-ning the uncertainty to the markets and tothe public all pose significant challenges forthe central bank.

Let me now turn to uncertainty aboutthe output gap. In its simplest form, thistype of uncertainty involves additive uncer-tainty, and the policy prescription from thestandard framework with a quadratic lossfunction is to make one’s best guess of out-put potential and to ignore the uncertaintysurrounding this guess. In practice, this canbe difficult, as I have discussed above. Butthere are also other factors that complicateboth theory and practice. As Orphanides(1999) has shown, errors in the estimates ofthe output gap can be highly autocorrela-ted, and this can lead to major errors inpolicy-making. Overestimates of produc-tivity growth and underestimates of thedegree of tightness in labour markets in theearly to mid–1970s were important ele-ments in the outbreak of inflation in a num-ber of countries. In addition, the use of fil-ters to estimate potential output is back-ward looking, and there can therefore bemajor problems at the end of the sample.Moreover, parameter estimates of such keyelasticities as the slope of the (inflation-aug-mented) Phillips curve are conditional onthe measures of the output gap. So, if thereis uncertainty about the output gap, there isalso uncertainty about these elasticities,which introduces multiplicative uncertain-

43

expectations were not well anchored, thiscould prove to be a recipe for actions being“too little, too late.” However, following thedeleterious experience of the “great inflati-on,” central banks have recognized theimportance of pre-emptive action to pre-vent inflation from re-emerging. Pre-emp-tive action is especially important wheninflation expectations are not well anchoredand can easily be unhinged. under such cir-cumstances, even a temporary demandshock could result in a rise in inflationexpectations because of past experienceand the uncertainty regarding the durationof the shock.Thus, it could affect the rate ofinflation.

In recent years, an important develop-ment that makes it somewhat easier for thecentral bank to cope with the kind ofdemand shocks that I have been describinghas been the success of policies aimed atachieving low inflation. The increased cre-dibility of these policies has provided ananchor for inflation expectations. In manycountries with inflation targets, longer-term inflation expectations have come tosettle on the centre of the inflation-controltarget range. This has allowed the centralbank an increased margin for manoeuvrewhen responding to demand shocks.Whereas, in the past, an upward movementin prices resulting from a demand shockwould have fed into inflation expectationsfairly rapidly, in today’s circumstances cen-tral banks have the luxury of taking more ofa wait-and-see attitude because the shock-induced increase in prices would not imme-diately affect inflation expectations. Put

another way, the reduction in both themarket’s and the public’s uncertainty aboutthe central bank’s commitment to contai-ning inflation has anchored longer-runinflation expectations. This has made iteasier for a central bank to deal with itsown uncertainty about whether a demandshock is transitory or longer-lasting bywaiting before taking policy action; i.e., itneed not be quite so pre-emptive. Thatsaid, waiting too long before acting couldbe risky (especially if the shock occurredwhen the rate of inflation was already at orabove its target) since credibility, and theassociated anchor for inflation expectations,could be lost if the central bank was per-ceived not to be reacting appropriately tothe inflation pressures from a longer-lasting demand shock.

A similar story can be told about sup-ply shocks, including the price-level increa-ses resulting from a currency depreciation.In the past, with uncertainty about the cen-tral bank’s commitment to keeping inflationlow and, hence, little or no anchor for infla-tion expectations, an increase in the pricelevel induced by a supply shock or by a cur-rency depreciation would rapidly feed intoinflation expectations. Hence, the centralbank would have to react quickly to such ashock to prevent, or at least minimize, theextent of a resulting wage-price spiral. Inthe 1990s, with inflation expectations morefirmly anchored, the public and the marketsare better able to differentiate betweenprice-level movements and inflation pressu-res, with the former requiring less of amonetary policy response than the latter.

42

in the forecast process by adding to the tra-ditional staff forecast of output and inflation(based on a forecasting model adjusted toincorporate the judgment of the staff) acompanion projection that is based onmovements of the financial variables (alsoadjusted for staff judgment). In addition tothe implications that movements in themoney and credit aggregates have for out-put and inflation, this companion projec-tion includes an assessment of the implica-tions of such financial measures as the termspread and the spread between conventio-nal and indexed bonds. The weight thatpolicy-makers will place on the variousinputs into the forecasting process willclearly depend on the track record that theygenerate over time in forecasting futuredevelopments. We are also examiningalternative reaction functions to see howrobust they are across different models.

I would now like to discuss how theexchange rate enters the policy process inCanada – another area where uncertaintyseems to have played an especially impor-tant role. A few years ago, the Bank ofCanada developed the concept of the mone-tary conditions index (or MCI) as a policyguide for a central bank in a small open eco-nomy. This concept was intended to captu-re in one measure both of the channelsthrough which monetary policy actionsaffect the economy; namely, interest ratesand the exchange rate.Thus, for example, ifa 25-basis-point increase in the centralbank’s benchmark interest rate led to a sig-nificant appreciation in the value of the cur-rency, this would imply much more tigh-

tening overall than if the value of the cur-rency remained unchanged or appreciatedonly a little in response to the policy action.Or, to put it slightly differently, the size ofan interest rate increase required to achievea desired amount of tightening in monetaryconditions would depend on the extent ofthe currency appreciation that accompaniedthe interest rate increase.

One implication of this approach isthat an exchange rate movement resultingsolely from portfolio adjustments on thepart of international or domestic investorswould require an offsetting interest rateadjustment to keep monetary conditionsunchanged. Now, it was made very clear inthe original analysis that other kinds ofshocks that affected the exchange rate, suchas a terms-of-trade shock, would require adifferent kind of response. For example, asignificant decline in the prices of rawmaterials, such as Canada experiencedduring the Asian crisis, would lead to both aweaker economy and a depreciation of theCanadian dollar. In this case, the currencydepreciation would be appropriate for theweakening economic and inflation situati-on, and there would be no reason to adjustinterest rates in an offsetting manner.

Use of the MCI as a policy guide,while appealing, had two basic difficultiesassociated with it. First, the markets startedto treat all exchange rate movements asportfolio shocks and therefore came toexpect an offsetting interest rate adjust-ment every time there was a movement inthe exchange rate, whether or not such an

45

ty. And the standard Brainard result is thatmultiplicative uncertainty calls for a morecautious response than would otherwisehave been the case.

The current situation facing Canadainvolves output near or at measured capaci-ty. But there is considerable uncertainty asto whether the conventional measure ofcapacity reflects reality, given all the policychanges and restructuring of the 1990s, andgiven the recent U.S. experience in whichunemployment moved well below traditio-nal measures of NAIRU without, thus far,leading to significant inflation pressures. Inassessing future developments, the Bank ofCanada’s response to this uncertainty hasbeen to put less emphasis on the measuredoutput gap and more emphasis on other lea-ding indicators of inflation.

We can interpret the U.S. experienceand the recent Canadian experience as aform of “probing,” in which the centralbank acts in a less pre-emptive manner thanotherwise when responding to a shock thatwould seem to test the limits of previousestimates of capacity. Nonetheless, in suchcircumstances, the central bank must beespecially vigilant for signs of pressures oncapacity (e.g., delivery delays or bott-lenecks) and pressures on inflation (e.g.,wage settlements or acceleration of moneygrowth).

More generally, multiplicative uncer-tainty applies to virtually all coefficients inestimated models. From a central bank’spoint of view, the most important is proba-

bly the considerable uncertainty attached toevery element of the transmission mecha-nism.This includes the effect of the centralbank’s action in changing the overnightinterest rate on money market rates, onlonger-term rates, on administered rates,and on the exchange rate, as well as theeffect of changes in interest rates and theexchange rate on spending, and the effect ofchanges in spending on the future rate ofinflation and on expected inflation.At everystage of the process there is considerableuncertainty, and at every stage, changes inexpectations play a critical role in determi-ning the outcome. In line with the literatu-re, this type of uncertainty leads the centralbank to be somewhat more cautious thanwould otherwise have been the case whentaking action, and is probably one of the fac-tors that leads to the “successive approxi-mation” approach to decision-making.

Let me now turn to model uncertain-ty. One important way in which this type ofuncertainty fits into the decision-makingprocess relates to the role that monetaryaggregates play in the transmission mecha-nism. Are they simply passive elements in aworld in which financial prices are the keydrivers, or do they play an active role in thetransmission of central bank actions todemand and inflation? At the Bank ofCanada, there are competing views on thisissue and on how uncertainty about the cor-rect model should influence the analysisleading up to policy decisions. In responseto this particular uncertainty about the cor-rect model, the Bank of Canada recentlyformalized the role of monetary aggregates

44

these conditional statements as a form ofcommitment. Following discussions withmarket participants, the Bank decided tofocus its discussion of future developmentson aggregate demand and inflation pressu-res and to leave it to the financial markets todraw their own conclusions about the likelyfuture path of monetary conditions.

This issue of possible limits to transpa-rency deserves further attention from aca-demics and policy-makers. And the bestway to address it would seem to be in thecontext of models that deal with the effectof uncertainty both on optimal ways ofmaking policy and on the best ways of com-municating with markets and with thepublic.

ReferencesBall, L., 1999. Policy rules for open economies.

In:Taylor, J. (Ed.), Monetary Policy Rules.

University of Chicago Press, Chicago, 127-156.

Bank of Canada.,1999a.Monetary Policy Report

(May).

Bank of Canada.,1999b.Monetary Policy Report

(Nov).

Duguay, P., Poloz, S., 1994. Role of economic

projections in Canadian monetary policy forma-

tion. Canadian Public Policy, June, 189-199.

Engert,W.,Selody,J.,1998.Uncertainty and mul-

tiple paradigms of the transmission mechanism.

Bank of Canada Working Paper 98-7.

Freedman, C., 1994.The use of indicators and of

the monetary conditions index in Canada.

In:Baliño,T.,Cottarelli,C. (Eds.), Frameworks for

Monetary Stability: Policy Issues and Country

Experiences. International Monetary Fund,

Washington, 458-476.

Freedman, C., 1995.The role of monetary condi-

tions and the monetary conditions index in the

conduct of policy. Bank of Canada Review

(Autumn), 53-59.

Freedman, C., 1996.What operating procedures

should be adopted to maintain price stability? –

Practical issues. In: Achieving Price Stability,

Federal Reserve Bank of Kansas City,Kansas City,

Missouri, 241-285.

Freedman, C., Longworth, D., 1995.The role of

the staff economic projection in conducting

monetary policy. In: Haldane, A. (Ed.),Targeting

Inflation. Bank of England, London, 101-112.

Laidler, D., 1999. The quantity of money and

monetary policy. Bank of Canada Working Paper

99-5.

Orphanides, A., 1999. The quest for prosperity

without inflation. Presented at the conference on

monetary policy-making under uncertainty,

European Central Bank, Frankfurt.

Smets,L.,1997.Financial asset prices and mone-

tary policy: theory and evidence. In: Lowe, P.

(Ed.), Monetary and Inflation Targeting. Reserve

Bank of Australia, Sydney, 212-237.

Srour,G.,1999. Inflation targeting under uncer-

tainty. Bank of Canada Technical Report 85.

Svensson, L., 1997. Inflation forecast targeting:

Implementing and monitoring inflation targets.

European Economic Review (41)6, 1111-1146.

Thiessen, G., 1995. Uncertainty and the trans-

mission of monetary policy. Bank of Canada

Review(Summer), 41-58.

Winkler, B., 1999.Which kind of transparency?

On the need for clarity in monetary policy mak-

ing. Presented at the conference on monetary

policy-making under uncertainty. European

Central Bank, Frankfurt.

47

adjustment was appropriate. Second, andthis difficulty faces all central banks in afloating exchange rate regime, the centralbank itself had to make a judgment on thesource of the shock to the exchange rateand the likely persistence of the shock inorder to decide on the appropriate response.

There are, of course, circumstances inwhich the source of the shock is evident.Terms-of-trade shocks or asynchronousbusiness cycle movements in Canada andthe United States are obvious examples.But when the origin of an exchange ratemovement is uncertain, how should thecentral bank react? Should such a deprecia-tion of the Canadian dollar be interpretedas the result of the market anticipating afuture weakening in raw materials prices ora slowing in the growth of Canadian dome-stic demand, or simply as the outcome of aportfolio readjustment by investors that isunrelated to the fundamental factors thatinfluence economic developments? An off-setting interest rate response would beappropriate in the portfolio-readjustmentcase but not in the case of a shock to funda-mentals (or what could be termed a “real”shock).

In the first half of the 1990s, the port-folio shock was the more prominent sourceof shocks to the Canadian dollar, while inthe latter part of the decade real shocks see-med to predominate. Therefore, our viewof the “default” interpretation for cases inwhich the source of the shock was uncer-tain, which is often the case, moved awayfrom a portfolio adjustment towards a real

shock. Indeed, Smets (1997) has arguedthat the reason Canada chose to use theMCI (which is particularly helpful in explai-ning the response to a portfolio shock) andAustralia did not, is that the typical shockaffecting the Canadian dollar at the time theMCI was adopted was a portfolio shock andthe typical shock affecting the Australiandollar was a real shock. And with theincreased importance of real shocks for theCanadian dollar in the later 1990s, the roleof the MCI has diminished in Canada.

There is one other issue relating tomonetary policy-making under uncertaintyto which I would briefly like to draw atten-tion and that is the notion that there mightbe limits to transparency. In practicalterms, this shows up in debates among cen-tral banks regarding the publication of theirforecasts. More broadly, one can questionthe extent to which central banks shouldexplicitly or implicitly signal either theirforthcoming actions or their tendencieswith regard to official interest rates. Forexample, the recent U.S. discussion aboutthe usefulness of the Fed’s announcement ofa “bias” focused on whether it helps marketsunderstand the Fed’s stance or simply cau-ses confusion. In Canada, conditional state-ments by the Bank of Canada about thefuture path of monetary conditions made inthe spring of 1998 ended up being under-mined by the impact on Canada of the sur-prising size and degree of contagion asso-ciated with the Asian crisis. Markets becameunsettled in the aftermath of these develop-ments. Part of the problem seems to havearisen from the market’s interpretation of

46

4948

Leonardo Leiderman,Research Department

Bank of Israel

Some Lessons from Israel

bution to this conference. Using an exten-ded version of Svensson’s small-scalemacroeconomic model of inflation targets,the author shows that parameter uncertain-ty acts to dampen current policy responses,much in the flavor of Brainard’s result. Tothe contrary, however, uncertainty aboutthe degree of inflation persistence results inmore aggressive policy responses thanunder certainty equivalence.This is the casebecause as the dynamics of inflation becomemore uncertain, strong policy actions canhelp reduce the uncertainty about the futu-re course of inflation and can shift inflationcloser to target. It turns out that, as indica-ted by the author, previous analytical workof Craine, Sargent, and Onatski and Stockproduced similar results about the policyeffects of uncertainty about the dynamics ofthe economy.

In reality, policymakers face of mix ofthese and other sources of uncertainty.While it seems that in many circumstancesin Israel – especially at relatively tranquiltimes--Brainard-type considerations haveled, in fact, to relatively smooth interestrate policy responses in reaction to shocks,policy in more turbulent recent inflationaryepisodes provides substantial support tothese relatively new ideas.

Before we turn to some details onthose episodes, in Israel’s case, that seem tosupport non-Brainard type reactions bypolicymakers, it is well to briefly describethe current monetary policy regime.Broadly speaking, inflation targeting hasplayed a key role in the disinflation in the

1990s. Following the well known inflationstabilization program of mid 1985, therehas been a double-digit rate of inflation inthe range from 10 to 20 percent in the late1980s and early 1990s, and recently inflati-on has been reduced to about 3-4 percentper year. The inflation target for 1999 was4 percent and the rate of inflation wasactually much lower than that (i.e., about1.3 percent), and the targets for the years2000 and 2001 have been set as the range of3 to 4 percent. For the first time since the1960s, there is a good chance that Israel’sinflation rate in the near future will not dif-fer much from that of advanced westerneconomies. While policymakers have beenoperating under inflation targets since1992, there has been a considerable degreeof ambiguity about the nature of these tar-gets, about the conduct of monetary policy,and about the fact that the inflation targethas also key implications for the stance offiscal policy and wage policy.A key difficul-ty has been the coexistence of inflation tar-gets with nominal exchange-rate targeting,in the form of a currency band, and the lackof a well specified preference orderingbetween the inflation target and the nomi-nal exchange rate target . Another difficul-ty has emerged in the mid 1990s, whenthere was a considerable degree of mone-tary policy overburdening aimed at offset-ting, at least partially, the inflationaryimpacts of a relatively expansionary fiscalpolicy that was not compatible with theinflation target. If anything, these conditi-ons have prompted the central bank tomaintain a more ‘conservative’ stance thanotherwise, in order to build up anti-

51

In the comments that follow, I elabora-te on what, in my view, is the main lesson onthe impact of uncertainty on monetarypolicy-making based on Israel’s experiencewith inflation targets in the 1990s. Thismain lesson is as follows: in an economywhere there is considerable uncertaintyabout the persistence of inflation shocks,and in which there is only partial credibili-ty of monetary policy in its attempt toachieve inflation targets – both these phe-nomena reflecting a history of high andvolatile inflation-relatively aggressivepolicy responses to deviations of inflationfrom target can have – especially whenunanticipated – a stronger and more effec-tive impact in bringing inflation back to thetarget, than smaller and gradual policyactions. I have no general theorem to sup-port this proposition at this time. However,it is claimed below that a plausible interpre-tation of the evidence, points in that direc-tion.

In spite of recent rapid convergence tosingle-digit rates of inflation, many keytransmission mechanisms that were intro-duced previously in Israel under double-and triple-digit inflation are still here.Leading examples are the existence ofwidespread wage- and financial-marketsindexation to the consumer price index andto the exchange rate, and a relatively quickpassthrough from nominal exchange ratesto prices. In addition, the history of highand volatile inflation left its mark, in a con-siderable degree of ambiguity, about thecompatibility of fiscal policy and wagepolicy with the inflation targets set by

government, and about the degree of com-mitment of monetary policy to achievingthese targets.Taken together, these featuresformed some of the key ‘initial’ conditionsfor disinflation in the 1990s, and as such,they play a central role in assessing theimpact of uncertainty on monetary policy-making in Israel, or in any other countrywhere present economic institutions andthe credibility of anti-inflation policies arestill influenced by a history of high inflation.

As also indicated by other participantsin this panel, and as documented in severalof the research contributions in this confe-rence, economic theory does not yieldunambiguous implications regarding theeffects of uncertainty on monetary policymaking. Brainard’s (1967) original contri-bution showed that uncertainty about theimpact of interest rate changes on inflationand output gives rise to more cautious, andless aggressive, monetary policy responsesthan would be the case under certaintyequivalence. Yet more recent analysis, byvarious authors, has questioned this resultand has pointed to a-priori ambiguity aboutthe impact of uncertainty. That is, onceBrainard’s one-period framework is exten-ded to a multi period dynamic setup, andonce various forms of uncertainty are con-sidered, various relations may emerge,including that which suggests that uncer-tainty should lead to more aggressivemonetary policy responses quite in contrastto Brainard’s result.

That this is the case has been stressede.g. by Sodertrom’s (1999) research contri-

50

end of the year. It seems that both theseunanticipated relatively aggressive policyactions had a major impact in reducing therate of inflation in later months.Subsequently, in a matter of four to fivemonths, there was a drop of 500 basispoints in inflation expectations, well intothe target range. In 1995, three consecutiveraises in the central bank rate, of 80 basispoints in May, 50 basis points in June, and150 basis points in July (the latter order ofmagnitude came as a surprise to the mar-kets) contributed to bring about a drop ininflation expectations which reached below10 percent at the end of that year.

The third illustrative episode has to dowith the period immediately following theRussian crisis and the LTCM failure in thesecond half of 1998. In Israel, the impact ofthe reduction in overall capital inflows toemerging economies resulted in a deprecia-tion of about 15 percent of the domesticcurrency against the US dollar, from Augustto October of that year.

The main concern for monetary policywas that what in other economies couldhave been a one-time jump in the nominalexchange rate and in the price level, couldbecome in Israel a permanent rise in therate of inflation. In the high-inflation past,currency devaluations had typically a strongimpact on inflation expectations and on therate of inflation. The quick passthroughfrom the exchange rate to prices, and to thelevel of inflation, reflected the existence ofwidespread wage- and financial indexationmechanisms, together with a high degree ofmonetary accommodation, all of whichgave rise to strong inflation persistence.Although these elements of the past wereeliminated to a great extent during inflationtargeting in the 1990s, the size of exchangerate depreciation this time – especially thefact that in about three days, from October6 to 8, 1998, there was a nominal exchangerate depreciation of about 10 percent –probably brought back some of the memo-ries from the earlier high-inflation period.That this is probably the case, is evident

53

inflation reputation and credibility and toensure the achievement of the targets, andthis more ‘conservative’ stance showed upin several quite aggressive policy responsesto potentially destabilizing shocks.

To illustrate this, I rely on three salientepisodes in which there were growing risksthat inflation might end up well above tar-get: in late 1994, mid 1996, and late 1998.These appear in Figure 1, which depicts theevolution of yearly inflation, market-based,one-year-ahead inflation expectations(derived from the yields on CPI-indexedbonds and non indexed bonds), and inflati-on targets. In the first two cases, a combi-nation of expansionary fiscal policy andother factors created inflationary pressureswhich resulted in within-the-year markeddeviations of forward-looking inflationexpectations from the inflation target.Accordingly, in the last quarter of 1994 –i.e., the first episode – inflation expectati-ons (as well as inflation forecasts) reachedthe high level of about 14 percent from a

level of about 7 percent in the first quarterof the year, while the underlying inflationtarget was 8 percent for the whole year.Similarly, in the second episode, toward themiddle of 1996 inflation expectations acce-lerated to a level of about 13 percent whilethe official target was the range 8-10 per-cent.

What was the monetary policy res-ponse to these marked deviations of inflati-on from target? It turns out that throughoutthe year 1994 there were gradual increasesin the central bank interest rate, in respon-se to a gradual increase in inflation expecta-tions over the year. The central bank ratewas raised from about 10 percent at thestart of the year to 12.5 percent in August.Yet, since these measures did not succeed inreducing the acceleration of inflation and ofinflation expectations, the central bank – ina somewhat unexpected action – raised theinterest rate in October by 150 basis points,and then again in December by the sameamount, to reach a level of 17 percent at the

52

1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000

25

20

15

10

5

0

% Annually

Israel – Actual Inflation, Targets and Inflation Expectations

Inflation targeting regime

Episode 1Episode 2

Episode 3Inflation expectations

CPI change,year over year

Indicates December of each year.Note: Inflation expectations are derived from capital market data on indexed and nonindexed bonds.

1992Target

1993Target

1994Target

1995Target

1996Target 1997

Target

1998Target

1999Target 2000-01

Target

Israel – Dollar Exchange Rate and 12 Month Inflation Expectations

17.3

16.3

13.3

11.3

9.3

7.3

5.3

3.3

% 4.5

4.4

4.3

4.2

4.1

4

3.9

3.8

3.7

3.68/98 9/98 10/98 11/98 12/98 1/99 2/99 3/99

X% Indicates a change in Bol interest rate by X%.

NIS

Russian crisis

Expectations

Dollar

LTCM

+2.0% +2.0%

-0.5%-0.5%

(Daily data)

ween uncertainty, credibility, and policyresponses is central to current monetarypolicy across many countries and regions,including that of the ECB, and is certainlyworthy of future theoretical and empiricalresearch.

55

from the daily data in Figure 2, where it canbe seen that the sharp currency depreciati-on was accompanied by a marked rise inone-year-ahead inflation expectations,which reached levels well above target.Thatis, marked exchange rate depreciation intwo to three days had a strong impact onone-year-ahead inflation expectations!

The central bank’s initial policy res-ponse to these events included no interven-tion in the foreign exchange market and noraise in the interest rate for September andOctober, on the ‘hope’ that under the infla-tion targeting regime this episode represen-ted more of a pricelevel jump than of apermanent rise in therate of inflation. How-ever, as inflation riskscould not be ruled out,and as inflation expec-tations exhibited someinertia at previouslevels, the central bankrate was raised at the end of October by200 basis points, and in the face of a lack ofa considerable downward adjustment ofinflation expectations the rate was increa-sed again by the same amount in two weeks,at an unexpected time, and this was percei-ved as ‘news’ by the markets. It was onlyafter this second, strong and unanticipatedrise in the central bank interest rate thatinflation expectations and the rate of inflati-on began to converge back to target.

Although these are episodes fromIsrael, the lessons that can be derived from

them have wide applicability.To generalize,in economies where there is considerableuncertainty about the persistence of inflati-on shocks – especially because of a historyof high and volatile inflation, whose impacton economic agents and institutions lastsfor a long period – and in which there isonly partial credibility of monetary policyin its attempt to achieve inflation targets,relatively aggressive policy responses todeviations of inflation from target can have--especially when unanticipated – a strongerand more effective impact in bringing backinflation to the target, than smaller and gra-dual policy shifts. This is in line with the

theoretical work dis-cussed at the start ofthis discussion. Thisdoes not mean thatBrainard-type conside-rations have not influ-enced monetary policyin Israel. In fact, inmany occasions themonetary authority

opted for cautious and gradual responses towhat were relatively small shocks in tran-quil circumstances, in an attempt to learnmore, over time, about the nature of theshocks and their effects on the economy.Yet, when major risks to the inflation targe-ting regime emerged – of a nature quiteanalogue to what a speculative attack doesto an exchange rate regime – more aggres-sive policies had an effective impact in avoi-ding transmission of one-time shocks to apermanent rise in the rate of inflation, andin strengthening the anti-inflation credibili-ty of the central bank. This relation, bet-

54

5756

Lucas Papademos,Governor, Bank of Greece

Central Bank Strategy and Credibility

ned, or by stabilizingthe exchange rateagainst the currency ofa low inflation country.The popularity of thesestrategies has variedover space and time.There are also other,more synthetic, appro-aches which involve elements or variationsof the above, such as the one pursued by theECB.

We should neither expect nor perhapseven aim at reaching a consensus on theappropriate strategy which is likely to besuperior under all “circumstances”, that is,for all countries independently of theirmarket or institutional structure and of thestate of their economy. This is because theoptimal strategy partly depends on the eco-nomic, financial and institutional structure,the likely stance of other policies, initialconditions, as well as the dominant sourcesof uncertainty facing a central bank over therelevant time horizon in which it formula-tes and implements policy. Of course, someof these determining factors may not beexogenous and may be altered by policiesand institutional changes to improve theeffectiveness of monetary policy. Let meillustrate this point with reference to theGreek experience.

Over the last ten years, the Bank ofGreece has adopted three different strate-gies in its attempt to disinflate the economyand achieve the necessary inflation conver-gence for joining EMU. During this period

inflation declined from an average rate ofaround 20% in the late eighties to about 2%today. In the late eighties and until 1993,the Bank employed a monetary targetingstrategy. The target range for broad moneygrowth was chosen to be consistent withthe inflation objective and projected realGDP growth.The choice of a monetary tar-geting strategy was justified by: (i) a fairlystable relationship between broad moneygrowth and nominal GDP growth, and (ii)the ability of the Bank to control reasonab-ly well, though not perfectly, the evolutionof the money stock.These two prerequisitesfor effective monetary targeting were partlymet because of the financial structure pre-vailing at the time: a banking system notfully liberalized, a relatively underdevelo-ped financial sector, and extensive capitalcontrols.

But, the adoption of monetary targe-ting was also a consequence of additionalconsiderations. Although the Bank had ahigh degree of operational autonomy to setand achieve intermediate monetary objec-tives, it had to accept the government’sinflation objective and, more generally, theorientation of monetary policy had to bebroadly in line with the overall objectives of

59

The theme of this conference might beinterpreted by some as suggesting that cen-tral bankers occasionally have the luxury offormulating and implementing policyunder conditions of certainty. Of course,the only certainty in central banking is theuncertainty surrounding it. In fact, uncer-tainty is the salient feature of the scienceand art of monetary policy-making. Thereare several types of uncertainty which com-plicate a central banker’s task:

• uncertainty associated with the impact ofexogenous variables or of non-monetarypolicies on the economy;

• uncertainty regarding market behaviouras well as the impact of technological andinstitutional change;

• uncertainty relating to the role and natureof expectations of future developmentsand policies, which crucially influence the transmission of monetary policy to theeconomy; and, consequently,

• uncertainty about market perceptions of the effectiveness of the central bank inperforming its monetary policy function.

I would like to focus on two related aspec-ts of this theme:• the implications of the nature and relative

magnitude of different types of uncertaintyfor the optimal choice of monetary policy strategy (which should be distinguished from the concept of a monetary policy rule);

• the importance of credibility in enhancing the effectiveness of monetary policy and the role of the institutional framework (for example, central bank independenceand accountability) in increasing credibility.

Although I will illustrate my remarksby drawing upon the experience of Greece,I believe that the general conclusions arenot country-specific.

In recent years, the monetary policystrategy adopted by the majority of centralbanks has involved some form of inflationtargeting, in the sense that the primarypolicy objective is to maintain price stabili-ty or reduce inflation, depending uponinitial conditions. This is fundamentally theconsequence of a widespread consensusamong economists, as well as among policy-makers, that long-term GDP growth cannotbe influenced significantly by monetarypolicy and that there is no trade-off bet-ween inflation and unemployment in thelong run. Furthermore, although an inflati-on-unemployment trade-off may exist inthe short run, theory and evidence suggestthat it is likely to be rather unstable so thatit cannot be exploited in a systematic wayby monetary policy. These propositions,regarding the nature of the inflation-unem-ployment trade-off in both the short andthe long run, are supported by the Greekexperience over the past twenty years.

The consensus which has emergedregarding the primacy of price stability asthe monetary policy objective has not beenaccompanied by a consensus on the appro-priate strategy for achieving it. Severalapproaches have been adopted, of which themost common are: direct targeting of futu-re (or forecasted) inflation, and indirect tar-geting of inflation, either by aiming to con-trol the growth rate of money suitably defi-

58

deficit widened. Uncertainty about thesustainability of the exchange rate targetand “rational” expectations about the needto maintain exchange rate stability withinthe ERM for at least two years before joi-ning EMU precipitated the entry of thedrachma into the ERM in March 1998.

The participation of the drachma inthe ERM led to a change not only in theinstitutional framework but also in themonetary policy strategy. Following ERMentry, monetary policy, although payingattention to the volatility of the exchangerate, has focused more on achieving thefinal policy objective of price stability, asexplicitly specified by the law which for-mally granted independence to the Bank ofGreece in December 1997. Thus, policyinstruments have been adjusting relativelymore directly to the future path of inflation,which is forecast by employing various eco-nomic and financial indicators, includingthe output gap, as well as a monitoring rangefor the rate of growth of broad money. Thelatter serves not as an intermediate targetbut as an indicator of the monetary policystance because of the considerable uncer-tainty surrounding the size of the outputgap and the non-inflationary rate of unem-ployment. The main reason, however,underlying the adoption of such a strategywas another type of uncertainty facingpolicy-makers. It related to the fact that thedrachma’s ERM entry was accompanied bya devaluation, which entailed a “shock” tothe price level and adversely affectedexpectations about the future course ofinflation and the exchange rate. It was the-

refore deemed necessary to pursue a morerestrictive monetary policy than thatimplied by targeting the central parity ofthe drachma within the ERM, in order tohelp absorb the inflationary effect of the“devaluation shock” and achieve price stabi-lity as quickly as possible, thus maintainingconfidence in the sustainability of the cen-tral parity.The greater flexibility for mone-tary policy allowed by the standard fluctua-tion band of + 15% in the ERM was there-fore considered useful, and was exploited.This strategy was also appropriate for con-trolling liquidity growth generated partlyby substantial capital inflows, induced byhigh domestic interest rates and a favoura-ble outlook for nominal convergence, andthe country’s eventual participation in theeuro area.

The effectiveness of the monetarypolicy, pursued in Greece in recent years, toreduce inflation and achieve price stability,reflects a number of factors. First, theadoption of the appropriate strategy giventhe structural and institutional constraintsand the dominant sources of uncertaintyaffecting the economy over the relevantperiod. Second, the flexible and resoluteuse of policy instruments to offset theimpact of external disturbances as well asinternally-generated inflationary pressures,favourably influencing expectations aboutthe prospects for price stability. Third, theenhanced credibility of monetary policy,following the successful implementation ofan exchange rate targeting strategy in pre-vious years, and the adoption of a new insti-tutional framework which not only granted

government policy. Fiscal and income poli-cies, however, were not always compatiblewith the inflation objective (either ex-anteor ex-post), while the system of wageindexation inhibited inflation control. Toput it simply, and in the present context,the cumulative uncertainty regarding therelationship between the ultimate inflationgoal and the monetary policy instrumentswas perceived to be far greater than thatconcerning the relationship between theseinstruments and the intermediate monetarytarget. Thus, direct inflation targeting wasnot only more difficult to implement, but italso involved risks for the credibility of thecentral bank within the prevailing institu-tional framework and in view of the uncer-tainty about the stance of economic poli-cies.

By mid-1994, the deregulation of thebanking system, financial innovation andthe full liberalization of capital movementshad impaired the stability of the relationshipbetween monetary aggregates and nominalGDP, and had undermined the ability of thecentral bank to control broad money effec-tively. Consequently, monetary targetingwas gradually abandoned and, betweenmid-1994 and spring 1998, the exchangerate of the drachma against a basket ofEuropean currencies became the interme-diate objective of monetary policy. Thechoice of this strategy was not only a conse-quence of the increased uncertainty associa-ted with the links between inflation, moneygrowth and interest rates. It was also influ-enced by uncertainties regarding themagnitude and evolution of the output gap

and, more importantly, expectations aboutthe stance of economic policies as well asabout the future behaviour of the exchangerate following the lifting of all capital con-trols. These uncertainties raised seriousdoubts about the effectiveness of directinflation targeting, and underlined theimportance of employing the exchange rate,both as a means of stabilizing expectations(about the exchange rate and the monetarypolicy stance) and as a device which couldhelp impose greater discipline on economicpolicy.The choice of the exchange rate as anintermediate objective was also affected bythe prospective participation of the drach-ma in the ERM, which was a preconditionfor joining EMU. In 1994, however, inflati-on was too high and fiscal policy too slackfor ERM participation to be considered aviable option.

The strategy of employing the exchan-ge rate as an intermediate target provedquite effective in reducing inflation, andincreased the credibility of monetarypolicy, which became more transparent.The credibility of the central bank wasenhanced because exchange rate targetswere met over a period of almost fouryears, during which the exchange rate wasdefended vigorously and successfully on anumber of occasions.Although this strategyincreased public confidence in the curren-cy, accelerated the pace of disinflation, andpromoted the fiscal consolidation process,it did not succeed in imposing the necessarydegree of discipline on the labour market.Consequently, international competitiven-ess was eroded and the current account

6160

The credibility of a central bank is animportant determinant of policy effectiven-ess because of its influence on marketexpectations, and thus on the dynamics ofthe transmission process. The Greek expe-rience suggests the following about thedeterminants of central bank credibility.The legal framework of a central bank,which safeguards its independence in achie-ving well-defined and feasible objectives forwhich it is accountable, is a necessary andessential determinant of its credibility.Nevertheless, the legal framework is notsufficient to ensure credibility and effectiveperformance. Another necessary ingredientis the implementation of economic andstructural policies which are compatiblewith the attainment of the monetary policyobjective. And, of course, a central bankmust be able to implement its strategy con-sistently, and use its policy instruments withflexibility and resolve. It must also employ a communications policy which can enhan-ce public understanding of its objectives,commitment, and strategy. In this way, infor-mational uncertainty can be reduced, andexpectations about the prospects for pricestability can be influenced so as to speed upthe attainment of the ultimate policy goal.

In the final analysis, however, the cre-dibility and effectiveness of monetarypolicy can only be established over time onthe basis of a central bank’s performance inan uncertain world, or, as a famous ancientGreek philosopher remarked more thantwo thousand years ago: “the proof of thepudding is in the eating”. Conferences suchas this one can help central bankers toimprove their recipes for their puddings.

63

the Bank independence in pursuing pricestability, but also made it accountable forachieving it. Fourth, an improved macroe-conomic policy mix, which resulted in fis-cal and labour market policies becomingprogressively more supportive of monetarypolicy. These factors contributed to redu-cing uncertainty about the orientation andimplementation of monetary policy, there-by increasing its effectiveness. I believe thatthe Greek experience is of broader relevan-ce and I would like to draw some generalconclusions regarding (i) the choice of aneffective monetary policy strategy, and (ii)the role and determinants of central bankcredibility in achieving the final policyobjective.

The review of the three monetarypolicy strategies pursued during differentphases of the disinflation process in Greeceillustrates how the choice of the appropria-te strategy depends upon the relativeimportance of different types of uncertain-ty facing policy-makers, including uncer-tainty regarding financial market behaviourand structure, the size of the output gap,the stance of non-monetary policies, andexpectations about the future path of keyvariables. However, the choice and effec-tiveness of a strategy depends on other fac-tors as well, such as the economy’s financi-al structure, its degree of openness, labour-market characteristics, and the institutionalframework of monetary policy.The variousstrategies pursued during the convergenceprocess of the Greek economy reflected, asI have argued above, the substantial changein these economic, structural, and institutio-

nal factors, as well as the relative magni-tude of different types of uncertainty.Theseconclusions were derived from the experi-ence of a given country in different periods.But, similar conclusions could be obtainedby comparing the alternative strategies pur-sued by different central banks in a givenperiod.

This leads me to express the view thatthe strategy of the ECB is the most appro-priate one for the euro area under the pre-sent circumstances. Pure monetary targe-ting could be subject to a high degree ofuncertainty in the new environment ofmonetary union, which may induce shifts inbehavioral patterns and/or the restructu-ring of the financial system. At the sametime, pure inflation targeting could be diffi-cult to implement credibly in view ofincreased uncertainty regarding the mone-tary transmission mechanism in the euroarea, and the expected impact of decentrali-sed national non-monetary policies. Thestrategy adopted by the ECB is likely to besuperior to these two alternatives. It mayappear relatively complex, but this does notimply that simpler alternatives would bemore effective in attaining the final objec-tive.We should not overlook the fact that achosen strategy, which is indeed superior toother possible alternatives, may still appearinsufficiently effective because of the influ-ence of economic and institutional factorswhich would also constrain the effectiven-ess of other monetary policy strategies aswell. These points should be taken intoaccount in the current debate on this issue.

62

6564

Alan S. BlinderPrinceton University and the Brookings Institution

Critical Issues for Modern Major Central Bankers

directly in the loss function – wholly apartfrom any other reasons for doing so.

Issue No. 2 is whether the central bankshould be an inflation targetter. This, ofcourse, is a relatively new issue. A decadeago, it was not even on the radar screen;nowadays, no modern central banker canavoid it. I must admit that my initial reac-tions to inflation targeting, while ViceChairman of the Fed, were quite negative.That was because I wrongly associated infla-tion targeting with placing zero weight onoutput stabilization – what Mervyn Kinghas called “inflation nuttery” – at a timewhen I was arguing strongly for just thereverse (Blinder, 1994). But I have learnedsomething since then. As Lars Svensson(1997) and others have shown, the weightthe central bank places on output stabiliza-tion maps directly into the speed withwhich the inflation target should be appro-ached. A central bank that places a highweight on output stabilization can be a gra-dualist inflation targetter. A second piece ofresearch that has made me more sympathe-tic to inflation targeting is Orphanides’ fin-ding, mentioned earlier, that large errors inestimating the output gap can lead (andhave led) to egregious errors in monetarypolicy. If an inflation targeting central bankstarts out underestimating (overestimating)the economy’s potential, it will see inflationfalling (rising), and therefore be induced toease (tighten) policy. That is an importantvirtue. A yet-newer literature, spawned byJapan’s problems with deflation and thezero bound on nominal interest rates, callsinto question the previous professional con-

sensus that inflation targeting is superior toprice-level targeting. In fact, this literaturesuggests that a price-level target might bepreferred when deflation is a danger. Thereason is simple: to get real interest ratesnegative when the zero bound on nominalrates is binding, the central bank needs toengender expectations of positive inflationeven though prices are falling. A credibleprice level target accomplishes that by pled-ging the central bank to offset episodes ofdeflation with subsequent periods of inflati-on, to get the price level back on its prede-termined path. (The trick, of course, is tomake the pledge credible.)More generally, central bankers must nowpay attention to an issue that their older bre-thren (they were all men then!) could safelyignore: the costs of deflation, which mosteconomists reckon to be greater than thecosts of inflation. Just like modern majorgenerals, modern central bankers must pre-pare to fight the next war rather than conti-nue fighting the last one.

Issue No. 3 pertains to transparency:How open should a modern central bankbe, and about what? Qualitatively, theanswer is simple: A modern central bankmust be a good deal more transparent thanits ancestors.There seems now to be some-thing approaching a consensus on this point– the consensus itself signifies a sea changein central banking attitudes. Among thelogical candidates for greater transparencyare the bank's ultimate goals (Issue No. 1above), its basic model of the economy(even if only informally), and its internalforecasts.

67

This conference is entitled “MonetaryPolicy-Making under Uncertainty.” But thattitle seems redundant to me – was thereever such a thing as monetary policymakingunder certainty? So I will adopt a differenttopic for this talk. With due apologies toGilbert and Sullivan, my focus is on what ittakes to be “the very model of a modernmajor central bank.” I will raise a host ofquestions, 15 in all, that would have to beanswered by a central bank starting do novoin the year 1999 – a situation that mayperhaps sound less than hypothetical here inFrankfurt. My emphasis is on how both thequestions and the answers differ from whatpeople might have thought ten or twentyyears ago – hence the adjective “modern.”The questions divide themselves into threecategories:1. issues of institutional design, or what mightbe called “setting up shop” (seven issues);2. tactics for operating in the markets (fourissues)3. issues pertaining to the bank's model of thetransmission mechanism (four issues).

I. Institutional DesignIssue No. 1, both literally and, I think,

figuratively, is the central bank's ultimategoal or goals for monetary policy – thearguments of its loss function. Almost allrecent academic research and thinking pre-sumes that the objective function of thecentral bank is some weighted average ofthe expected squared deviations of outputand inflation from their respective targets.But that raises several subsidiary questions:

(a) What are the weights? Just howmuch should the central bank care about

output (employment) deviations relative toinflation deviations? This choice is crucial,but underemphasized. It can exercise sub-stantial influence over actual policy decisi-ons. For example, it may be one of the keypoints of difference between the ECB andthe Fed today.

(b) Around what targets? The choice ofthe targets is probably more important thanthe choice of the weights.The inflation tar-get has been extensively examined both inacademic literature and in central bank dis-cussions. But what about the output target?Athanasios Orphanides' paper for this con-ference suggests that specifying an outputtarget in a sensible way can be quite difficultin real time.In addition, a deep question arises if theeconomy displays hysteresis: Does it theneven make sense to specify an output targeta priori, when doing so might lead the cen-tral bank to settle for a local optimum eventhough there may be a superior global opti-mum available? As a concrete example,think about how much worse off the UnitedStates (and, indeed, the entire world)would be today if the Fed had decided in1995 that the U.S. economy could notsustain an unemployment rate below 6% –and had acted on that belief.

(c) What about financial stability? Realcentral bankers care about more than justthe variances of inflation and output. Theyalso bear a responsibility for financial stabi-lity, which, while related to the other twogoals, is not entirely subsumed in them. Inmy view, concern over financial stability isthe best rationale there is for includingsomething like the change in interest rates

66

central banks’ unwillingness to intervenewith large amounts of money.

Issue No. 6 is whether the monetaryauthority should also regulate and supervi-se banks. This issue is very much up in theair right now.The U.K. has explicitly sepa-rated monetary policy from bank supervisi-on, as you know, and the ECB is not a banksupervisor. (But several central banks wit-hin the ESCB are.) In the U.S., we have justconcluded a multi-year turf war betweenthe Federal Reserve and the TreasuryDepartment over the Fed's role in banksupervision.Throughout, the Fed has stead-fastly insisted that the information it routi-nely receives in its supervisory role, is vitalto the performance of its monetary-policyduties. Is that true?My personal view is that the Fed has taken agrain of truth and greatly exaggerated itsimportance. Proprietary information thatthe central bank receives in bank examina-tions is of some, limited use in formulatingmonetary policy – and is on rare occasionsvery important. So, on balance, it is proba-bly better to have it than not. On the otherhand, a bank supervisor may sometimeshave to be a protector of banks and someti-mes a stern disciplinarian – and either

stance may conflict with monetary policy.In the United States, there is yet anotherconflict of interest, which is currentlyunder study by the General AccountingOffice:The Federal Reserve not only super-vises banks, it also sells them priced servi-ces in competition with private vendors ofthe same services.Finally, two other questions are worth rai-sing in this context. First, as the lines sepa-rating banks from other financial instituti-ons blur and disappear in the modernworld, must central banks that serve asbank supervisors be morphed into general-purpose financial supervisors – and do theyhave the expertise to do this broader job?Second, even if we decide that central banksshould be bank supervisors, why shouldthey also be bank regulators, that is, rule-makers (as the Federal Reserve is)?Shouldn't that function remain in the politi-cal domain?

Issue No. 7 is genuinely novel – centralbankers of a generation ago certainly didnot think about it. The question is this: Dovarious (actual and incipient) forms of elec-tronic money pose a threat to central banks?Two distinct sorts of threats can be imagi-ned; both arise from the possible erosion of

69

As many of you know, I have long been ahawk on transparency – on both economicand political grounds. Economically, I belie-ve greater transparency makes monetarypolicy easier by anchoring expectations bet-ter to the realities underlying policy.Financial markets that are better attuned tothe central bank's thinking are better able toanticipate its actions. And, anticipatorymovements in interest rates, if correct,shorten the lag in monetary policy – a lagthat has long bedeviled attempts to stabilizethe economy. Politically, democratic theorystrongly suggests that, in return for itsbroad grant of authority, the central bank isobliged to keep the public and their electedrepresentatives well informed.To do other-wise would be imperious. (Remember theetymology of that word!)

Issue No. 4 is rarely discussed, butmust be considered at the design stage:Should monetary policy decisions be madeby an individual or by a committee? I amcurrently engaged in some experimentallaboratory research at Princeton to test twohypotheses: that, compared to individuals,H1: committees react more slowly to thesame stimulus.H2: committees nonetheless make betterdecisions.It is a bit too early for definitive results, butthe early returns dispute H1 while suppor-ting H2. More generally, I want to take noteof a small academic literature that is develo-ping around the question of whether, andhow, monetary policy decisions made bycommittee differ from monetary policydecisions made by individuals.

The Federal Reserve offers an interesting,and apparently highly successful, hybridmodel. The Federal Open MarketCommittee (FOMC) in a formal sensemakes decisions by majority rule with arecorded vote. But, in fact, it is dominatedby its chairman. Much of the outside worldis watching to see whether the ECB willdevelop into an FOMC-style faux commit-tee, or into a genuine committee organizedon the “one person, one vote” principle(like the Bank of England’s Monetary PolicyCommittee).

Issue No. 5 is whether a central bankoperating in a floating exchange rate regimeshould forsake foreign currency interventi-on as a policy tool – even though theexchange rate is an important part of themonetary transmission mechanism.The conventional wisdom nowadays seemsto be that central banks should forget aboutintervention, mainly on the grounds thatsterilized intervention doesn't work. But Iwonder if this is always right. Certainly, for-eign exchange interventions that opposemajor market trends stand little chance ofsuccess; the old market wisdom, “don’tstand in front of a freight train,” applies tocentral banks as well. But there are timeswhen markets have no particular convictionabout which way the exchange rate will gonext, or are thin, or have lots of nervousshort-sellers.At such times, the markets aresusceptible to being pushed around (withinlimits) by the central bank – if it is willing tocommit substantial sums to the effort. Itcould be that the current consensus againststerilized intervention stems, in part, from

68

parent ways. Should monetary policy justproceed as if none of this had ever happen-ed? I think not. At a minimum, a moderncentral bank must make use of the informa-tion found in these new markets. For exam-ple, the Fed has for years used the Federalfunds futures market in Chicago as its pri-mary window into what the markets arethinking about future monetary policy. Inaddition, high-tech financial instrumentsalmost certainly affect the monetary trans-mission mechanism – especially the linka-ges from short rates to long rates and otherfinancial market prices.And, central bankersignore this at their peril. My own hunch –but it's just a hunch – is that derivatives haveshortened the lag in monetary policy.There are still more questions. Should amodern central bank operate in some ofthese more exotic markets, rather thanrestrict itself to conventional open-marketoperations in government securities? A con-servative central banker’s reflexive answeris no, and I must admit this is my own reac-tion to date. But perhaps the idea shouldnot be dismissed out of hand.After all, deri-vatives can enhance the power of the centralbank to push interest rates (or even exchan-ge rates) around, just as they do for privatemarket participants. A modern central ban-ker needs to give this issue serious thought.

Issue No. 10 pertains to what I call theBrainard (1967) conservatism principle: theidea that multiplier uncertainty should makethe central bank more conservative, in thesense of using its policy instrument lessvigorously. In Blinder (1998), I opinionedthat, while the conservatism principle is not

very robust mathematically, “My intuitiontells me that [it]... is more general – or atleast more wise – in the real world than themathematics will support.” (p. 12)This remark seems to have touched off a fairamount of quite interesting academic work,and I have been surprised at how little sup-port Brainard’s principle has received.There are, by now, a number of examples inwhich multiplier uncertainty, in conjunc-tion with something else, leads an optimi-zing central bank to vary its instrumentmore than it would under certainty. TheBrainard result is indeed fragile. Still, I findthese new anti-Brainard results both puzz-ling and troubling.Though my confidence inthe conclusion has been shaken by recentresearch, my gut still tells me that Brainardwas right in practice. In any case, it's cer-tainly an intellectual question that shouldengage modern central bankers.

Issue No. 11 is related:When a centralbank decides to change monetary policy,should it move its interest rate by large orsmall amounts? Under Alan Greenspan’sstewardship, the Federal Reserve has showna clear preference for frequent, small moves– usually 25 basis points. And who wouldargue that the Greenspan Fed has not beensuccessful? Yet, I suspect that this style ofpolicy is not what dynamic optimizationcalls for.Why not? The argument for largermoves is predicated, in part, on the unitroot in the inflation process: If inflation canrandom-walk away from you, the centralbank will want to make sure to step on thebrakes hard enough. But what if the unitroot in the inflation time series is a bypro-

71

the bank's monopoly over the issuance of themedium of exchange. The first threat is tocentral bank independence: If seignoragerevenue shrinks enough, the central bankwill become beholden to the legislature forits annual budget, and that could make itmore susceptible to political influence. Lossof seignorage revenue is probably not muchof a threat to the big three (Fed, ECB,BOJ), but it could be a more serious matterfor smaller central banks whose expensesmore nearly exhaust their revenue. Thesecond threat is to monetary policy itself: IfMicrosoft-money and the like come to beused for settlements on a grand scale, bankswill no longer need reserves at the centralbank for clearing purposes. Indeed, thebanking system might be bypassed entirelyif buyers and sellers settle accounts directlywith e-cash. Similarly, electronic transfersof all sorts make it increasingly easy forbanks to avoid the implicit tax on requiredreserves – as sweep accounts have beendoing in the U.S. for years. In combination,these two developments will weaken – andmay eventually even destroy – the mainlever that central banks have traditionallyused to manage their economies: controlover base money.What's a central banker todo? For now, I think, the answer is: nothing.But sometime, in the near future, thesehypothetical questions may become realones which modern central bankers will beforced to confront.

II. Tactics for Operating in theMoney Market

My next four questions relate to how thecentral bank operates in the financial markets.

Issue No. 8 is a broad question of stra-tegy rather than a narrow tactical one. If Imay be forgiven for indulging in stereoty-pes for a moment, some years ago, centralbankers saw their proper role as surprisingand bullying the markets. Central bankerswere (they thought) in control; marketswere meant to be pushed around. No lon-ger. Nowadays, a thoroughly modern cen-tral banker is more likely to respect marketsand keep them well informed. That is ahealthy development, but it can be takentoo far.As I emphasized in my Robbins Lectures(Blinder, 1998), central bankers are oftentempted to “follow the markets” – that is, todeliver the monetary policy the markets areexpecting or, indeed, demanding. At times,that might be precisely the right thing to do– especially if the bank has conditionedmarket expectations properly. But notalways. Many of us believe that marketstend to go to extremes, to overreact to sti-muli, and to be stunningly shortsighted. Agood monetary policymaker must succumbto none of these temptations.

Issue No. 9 concerns the implicationsof high-tech finance for the conduct ofmonetary policy. A host of questions formodern central bankers arise here. Howshould monetary policy adapt to the explo-sion of derivatives and financial exotica ofall kinds – instruments that central bankersnever dreamed of a decade or two ago?Some of these markets are extremely deepand liquid; some contain a great deal ofinformation; many of them create extreme-ly high leverage – sometimes in non-trans-

70

Issue No. 14 is a closely-related para-dox in the international arena. Uncoveredinterest parity is supposed to tie currentand expected future exchange rates to theinterest-rate differential between any twocountries. Yet it fails miserably as a foreca-ster of future exchange rates. Once again,everybody knows this, but no one (myselfincluded) seems to know what to do aboutit. Since interest parity is an essential link inthe monetary transmission mechanism foropen economies, and since all economies areopen, this is not only an intellectual embar-rassment but a major impediment to success-ful monetary policy. It must rank high onthe work list for modern central bankers.

Last, but certainly not least, I come toIssue No. 15: How does a central bank con-duct monetary policy in the absence of aPhillips curve it can trust? For years, I usedto gloat that the Federal Reserve had animportant advantage over the other G7central banks: We had a reliable statisticalPhillips curve to use, they did not. Butnowadays we all seem to be in the sameboat. As is well-known, the traditional U.S.Phillips curve, which worked so well fordecades, has been malfunctioning of late.Today the Fed finds itself up the creek with-out a Phillips-curve paddle, just like othercentral banks. This is a serious handicap.Given the long lags in monetary policy, it isgenerally agreed that the authorities needto conduct a “preemptive” monetary policy.That means moving on the basis of inflationforecasts. But, the collapse of the Phillipscurve leaves us without a reliable way toanticipate the impacts of economic activity

on inflation. And that, in turn, raises aserious intellectual question:When is it bet-ter to wait for an actual upturn in inflationrather to act preemptively, on the basis of aforecast? Both Orphanides’ paper and theBrainard uncertainty principle suggest thatthe current preference for preemption mayneed reexamination.

IV. in Conclusion So, that is my highly-selective list of 15 cri-tical issues. Rather than try to sum up, I willagain beg the indulgence of Gilbert andSullivan, and conclude in verse:

I am the very model of a modern centralbank for all.I’ve information national, financial, interna-tional.I know the Bank of England, and I quote theminutes of the Fed.I mimic every syllable that Alan Greenspanever said.I’m very well acquainted, too, with mattersmathematical.I understand equations, both the simple andquadratical.Of standard deviations, I am teeming with alot o’news,With many useful facts about the square ofthe hypotenuse.I’m very good at integral and differentialcalculus.My staff provides me models that are reallyquite miraculousIn short, in matters national, financial,international,I am the very model of a modern centralbank for all.

73

duct of a particular policy regime that letinflation ratchet up from the 1950s to the1960s to the 1970s? As more recent data areadded to the sample, the evidence for a unitroot weakens. In other words, an appro-priate monetary policy – one that approxi-mates a Taylor rule, say – can remove theunit root from the inflation process. In thatcase, a more gradualist approach to mone-tary policy might make sense. It's some-thing for a modern central banker to thinkabout.The ECB, according to current marketlore, prefers larger, less frequent moves –say, 50 basis points. But I caution you thatsuch a deduction is based on a rather thindata base – precisely two observations!

III. Questions about the MonetaryTransmission MechanismMy final four questions pertain to the cen-tral bank's model of the economy.

Issue No. 12 should be an easy one,although the ECB seems not to agree. Inthis case, I will state an answer rather thanpose a question: A modern central bankshould think of its overnight interest rate,not any monetary aggregate, as its principalpolicy instrument. My reason is simple andwell known. As Gerry Bouey, a formerGovernor of the Bank of Canada, aptly putit, “We didn’t abandon the monetary aggre-gates, they abandoned us.” With financialinnovation virtually certain to continue,and with the lines between banks and othertypes of financial institutions blurry andgetting blurrier, I see no reason to suspectthat this abandonment will end soon.A cen-

tral bank that relies on a monetary aggrega-te may trap itself in vestigial thinking – andmay therefore put its economy in harm’sway. Indeed, modern financial arrange-ments are rapidly eroding the primacy ofbanks, which are the source of the moneysupply as conventionally defined. Suchdevelopments scream out to central banksto stop focussing on the textbook link frombank reserves to bank lending to aggregatedemand. Instead, a modern central bankshould think of the main linkages in thetransmission mechanism as running from itspolicy rate to other interest rates and finan-cial prices (such as longer-term interestrates, exchange rates, and stock marketvalues), and then on to aggregate demand.The Ms are byproducts of this process, butof no great intrinsic interest.The next two issues follow directly fromthis point of view, and are vexing ones. Butsince they are also familiar, I will deal withthem briefly.

Issue No. 13 observes that the stan-dard model linking short- and long-terminterest rates – the so-called expectationstheory of the term structure – is deadwrong, in the sense that long rates are ter-rible predictors of future short rates. Thisfact seems to be well-known in academia, inthe markets, and in central banking circles.But, its resolution remains a mystery. Giventhe importance of long-term interest ratesto the monetary transmission mechanism,this may be the single most importantintellectual issue with which modern cen-tral bankers must grapple. I wish I could tellyou the answer, but I can't.

72

75

I know central banking history fromLombard Street to ECB.I speak in cryptic phrases whose intent israther hard to see.With repos, I can push the rates from floorto ceiling flawlessly.And seignorage enables me to prosper rat-her nice-a-lyMy monetary knowledge is extensive andadventury.It’s based on all the wisdom handed downacross the centuries.And so, in matters national, financial, inter-national,I am the very model of a modern centralbank for all.

BibliographyBall, Laurence, “Efficient Rules for Monetary

Policy,” NBER Working Paper No. 3952 (1997).

Blinder, Alan S., Central Banking in Theory and

Practice, Cambridge, MA: MIT Press (1998).

Blinder, Alan S., in Reducing unemployment:

Current issues and policy options: A symposium

sponsored by the Federal Reserve Bank of Kansas

City (1994).

Brainard, William, “Uncertainty and the

Effectiveness of Policy,” American Economic

Review 57 (May 1967), pp. 411-425.

Brayton, Flint, John M. Roberts, and John C.

Williams, “What’s Happened to the Phillips

Curve?,” Division of Research and Statistics,

Federal Reserve Board (1999).

Faust, Jon.“Whom Can We Trust to Run the Fed?

Theoretical Support for the Founders’ Views,”

Journal of Monetary Economics 37 (April

1996), pp.267-283.

Fischer, Stanley, “Modern Central Banking,” in

Forrest Capie, Charles Goodhart, Stanley Fischer

and Norbert Schnadt, The Future of Central

Banking, Cambridge, UK: Cambridge University

Press (1994).

Sibert, Anne, “Monetary Policy Committees:

Individual and Collective Reputations,” Birbeck

College (1999).

Svensson, Lars E.O.,“Inflation Forecast Targeting:

Implementing and Monitoring Inflation Targets,”

European Economic Review 41 (1997), pp.

1111-1146.

Svensson, Lars E.O., “Price Level Targeting vs.

Inflation Targeting,” Journal of Money,Credit and

Banking (1999).

Wolman, Alexander L., “Real Implications of the

Zero Bound on Nominal Interest Rates,”Working

Paper, Federal Reserve Bank of Richmond

(1998).

Woodford, Michael, “Optimal Monetary Policy

Intertia,” NBER Working Paper No. 7261

(1999).

74

Session Summaries

distributed lag of the observable variables,the estimate is updated only gradually.Therefore, even under discretion, the ob-served optimal policy will display consi-derable inertia as a consequence of partialinformation.The degree of inertia is higher,the noisier the current observations, andthe less the weight on current observationsis relative to previous estimates.

Svensson and Woodford illustrate theuse of the modified Kalman filter formulaefor deriving the optimal weights of indica-tor variables by analysing a simple New-Keynesian model ofmonetary policy, whichcomprises a forward-looking aggregate sup-ply relation and a for-ward-looking IS rela-tion. Potential outputis assumed to beunobservable, and thesupply shock is mea-sured only with error.The vector of obser-vable variables includes the forward-loo-king inflation rate which both affects anddepends on the current estimates of poten-tial output and the supply shock. By apply-ing the modified Kalman filter, Svenssonand Woodford show that the usefulness ofthe observable variables for optimal policy,i.e. their optimal indicator weight, dependson the noise surrounding their measure-ment. Due to the separation of estimationand control, however, the weight to be put,on the resulting estimate of potential out-put under an optimal policy is unaffected bythe amount of noise involved. Svensson

and Woodford therefore conclude that thelack of a more accurate estimate of the out-put gap itself is not a reason for policy to beless responsive to perceived fluctuations inthe output gap.

Summary of the DiscussionJose´ Vinãls’ discussion of the paper

mapped it from its technical insights to thepractical problems of monetary policy-making under uncertainty. He welcomedthe setting of the technical analysis addres-sing the real world complexities which arisedue to partial information about the state of

the economy, and thenon-resolvable link be-tween monetary policyactions and the for-ward-looking expec-tations of the privatesector. He appreciatedthe authors’ message tocentral bankers thatsuch circumstances

should not worry them. Given the pre-sumptions of the analysis, central bankersshould conclude that monetary policy canbe simplified. First, central bankers can actas if the partially observable variables wereknown (due to certainty equivalence and,thus, the separation of estimation and opti-misation). Second, optimal policy undercommitment is superior to discretionarypolicy, even in an environment of uncer-tainty about the state of the economy.Third, optimal policy, both under commit-ment and under discretion, is characterisedby inertia. This inertia may increase theeffectiveness of monetary policy actions by

77

Summary of the PaperThe paper by Svens-

son and Woodford aims atclarifying the principles ofderiving the optimalweights on indicators foroptimal policies in models with partialinformation about the state of the economyand forward-looking variables. As a chal-lenge for deriving the optimal weights, for-ward-looking variables complicate both theproblem of estimating the partially obser-vable state of the economy, as well as thedetermination of optimal policies. Thischallenge arises because forward-lookingvariables depend, per definition, on cur-rent expectations of future endogenousvariables and the current setting of thepolicy instrument. Current expectationsand the current instrument setting dependon the estimate of the current state of theeconomy. This state, in turn, depends onthe current realisation of the forward-loo-king variables, thereby causing a simulta-neity problem.

In addressing this simultaneity pro-blem, Svensson and Woodford first re-esta-blish the important result that, under sym-metric partial information, certainty equi-valence and the separation principle conti-nue to hold in the case of linear rational-expectations models and a quadratic lossfunction. Thus, the determination of opti-mal policies, given an estimate of the parti-ally observable state, and the estimation of

the state, can be treated asseparate problems andoptimal policies, as a func-tion of the estimate of thecurrent state of the eco-nomy, act as if the state

were observed.With regard to the problemof estimating the partially observable state,they show that the estimation problembecomes a simpler variant of the estimati-on problem with forward-looking variableswhich was previously solved in Pearlman,Currie and Levine (1986) by transformingthe original problem to a problem withoutforward-looking variables. In dealing withthe optimisation problem, they provide theoptimal policies under discretion as well asunder commitment, building on Pearlman(1992). The most fundamental differencewith respect to the discretion case is thatoptimal monetary policy under commit-ment depends on the history of the eco-nomy, and thus displays inertia.

For both the discretion case and thecommitment case, Svensson and Woodfordprovide some general results on how theobservable variables – the indicators – areused to estimate the current state of theeconomy. The solution to the estimationproblem with forward-looking variables isexpressed in terms of an appropriatelymodified Kalman filter in which the Kal-man gain matrix gives the optimal weightson the observable variables for estimatingthe state. Since the estimate of the state is a

76

Lars E.O. Svensson (Stockholm University and Princeton University) and

Michael Woodford (Princeton University)

Indicator Variables for Optimal Policy

implementing inflation targeting in modelswith forward-looking variables and partialinformation, Tabellini criticised the simplestructure of the New-Keynesian modelused for illustrating the results. Since infla-tion is a leading indicator of the output gapwithin this class of models, he questionedthe emphasis on the output gap as an infor-mation variable for monetary policy.Instead, referring to Galí and Gertler(1998), he suggested that marginal costswould be the more appropriate variable tofocus on in the model’s context.

ReferencesBarro, R.J. and D.B. Gordon (1983), “A Positive

Theory of Monetary Policy in a Natural Rate

Model”, Journal of Political Economy, 91, 589-

610.

Brainard, W. (1967), “Uncertainty and the

Effectiveness of Policy”, American Economic

Review Papers and Proceedings, 57, 411-425.

Galí, J. and M. Gertler (1998), “Inflation

Dynamics: A Structural Econometric Analysis”,

mimeo, New York University.

Pearlman, J.G. (1992), “Reputational and

Nonreputational Policies under Partial Infor-

mation”, Journal of Economic Dynamics and

Control, 16, 339-357.

Pearlman, J.G., D. Currie and P. Levine (1986),

“Rational Expectations Models with Partial

Information”, Economic Modelling, 3, 90-105.

79

exploiting the leverage of the expectationchannel. Last but not least, a modifiedKalman filter gives the optimal weights ofthe multiplicity of indicator variables formonetary policy which, in practice, mayoften provide conflicting information oneconomic developments.The Kalman gain’sdependence on the structure of the modeldemonstrates that it is crucial to know theproperties of the monetary transmissionmechanism in order to determine theweights of the available indicator variables.Therefore, a pure statistical horse race forthe importance of indicators must be inap-propriate. Vinãls thanked the authors fortheir valuable contribution but also stressedthat reality is even more complex than pre-sumed in the paper. A central bank, in fact,would have to rely on imprecisely estimatedbehavioural parameters which are assumedknown in the paper. In the context of para-meter uncertainty, however, certainty equi-valence does not hold any longer, and thecautious recommendation of Brainard (1967)to follow a more gradualist approach tomonetary policy would prevail. Moreover,the results by Svensson and Woodfordwould have to be reconciled with the recentfindings based on the analysis of Taylor-typepolicy rules incorporating a feedback of themonetary policy instrument to an impreci-sely measured output gap. In contrast to theconclusion by Svensson and Woodford,these latter findings recommend not res-ponding aggressively to fluctuations of theimprecise gap measure.

The discussion by Tabellini startedwith a brief review of Svensson’s previous

work on the analysis of strict versus flexibleinflation targeting in predominantly back-ward-looking models for which the princi-ples of certainty equivalence and the sepa-ration of estimation and optimisation arewell known. Against this background,Tabellini appreciated the extensions bySvensson and Woodford in re-establishingthese principles for models with partiallyobservable states in the presence of for-ward-looking variables, and for providingsimplified filtering formulae. With regardto the derivation of optimal policies underdiscretion, Tabellini referred to the simple,static Barro and Gordon (1983) frameworkwhere the socially optimal monetary policycan be enforced by a linear inflation con-tract, or by appointing a conservative cen-tral banker. He emphasised Svensson’s andWoodford’s finding by that in dynamicmodels with forward-looking variables, thesocially optimal monetary policy can simi-larly be enforced under discretion byappropriately adding lagged terms of theLagrange multiplier associated with the for-ward-looking variables to the central bank'sobjective function. The question remains,however, whether these terms can be sim-plified, and which practical implications thisfinding may have for the institutional designof monetary policy. Since the assumption ofsymmetric partial information may beunrealistic for monetary policy, he welco-med ongoing work by Svensson andWoodford which aims at extending theresults to the case of asymmetric informati-on for which certainty equivalence does nothold. Though sympathetic with the outco-me of the positive and normative analysis of

78

capture the true trade-off between inflationand unemployment at the self-confirmingequilibrium. Would it be possible for theauthorities to learn the natural rate hypo-thesis? Only in this case they would opti-mally set the systematic component of inflati-on to zero, leading the economy to the opti-mal Ramsey outcome. However, learning thenatural rate hypothesis would require suffi-cient variation in private inflation expectati-ons.This variation cannot arise, because in theself-confirming equilibrium, the governmentdoes not change the systematic component ofinflation, so that inflation expectations alsoremain constant.

Experimentation can be started by acertain unusual shock pattern, which indu-ces the authorities to randomize the syste-matic component of the inflation rate. Thisgenerates a data scatter that increases theslope of the estimated Phillips curve. Theoptimal reaction of the government is thento lower inflation, thus generating observa-tions that make the estimated curve stee-per. This process ends when the estimatedPhillips curve is vertical, so that the eco-nomy is at the Ramsey optimal outcome.However, the economy cannot stay thereforever.The government would soon disco-

ver and exploit the true short-run Phillipscurve, moving back to the time-consistentself-confirming equilibrium.

Summary of the DiscussionIn the first discussion, Ramon

Marimon criticized the specification of themodel perceived by the government, wherethe role of expectations is completelyignored. A more realistic setting would beone in which the authorities still use anapproximating model, but where a measureof inflation expectations enters as a regres-sor, in addition to the constant and to reali-zed inflation. Sargent’s setup would then bea particular case in which the coefficient oninflation expectations is set at zero. Underthis alternative model assumption, i) thegovernment would also react optimally tochanges in expectations; ii) the econometri-cian would be using a recursive algorithmto estimate a long-run Phillips curve; iii)under the natural rate hypothesis, the coef-ficients on realized inflation and on inflationexpectations would be identical.This morerealistic set-up might lead to different con-clusions. First of all, multiple self-confir-ming equilibria might arise. Second, analternative explanation of the US inflationpattern may emerge; Volker’s and

81

Summary of the PaperTom Sargent presented

a paper co-authored with In-Koo Cho on “Escaping NashInflation.” The paper aims atexplaining the observed pathin U.S. inflation, after WorldWar II. Following a decade of relatively sta-ble inflation, the US experienced an accele-ration in inflation in the late ‘60s and ‘70s, asharp stabilization in the ‘80s and a furtherslowdown in the ‘90s. Two alternativeapproaches have been used in the literatureto explain the data. The first argues thatmovements in observed inflation reflectedmovements in the time-consistent inflationrate, due to changing fundamentals in thebasic Kydland-Prescott model. The secondargues that, after the early 1980’s, themonetary authorities may have choseninflation below the time-consistent rate.

Building upon the literature on leastsquares learning, Sargent and Cho contri-butes to the second approach. In his modelfundamentals are fixed, while it is assumedthat:1) the authorities (imperfectly) control

inflation, and the private sector forecasts inflation optimally;

2) the true data generation process embo-dies an expectational Phillips curve;

3) the authorities dislike inflation and unem-ployment;

4) the authorities do not know the true data generating process, but use an approxi-

mating model, whereunemployment is regressedon a constant and on obser-ved inflation.

Previous works have shownthat an adaptive system

with least squares learning is driven by adeterministic dynamics, known as ‘meandynamics’, and that it converges to a self-confirming equilibrium.This is an equilibri-um concept that is suited for approximatingmodels.While it differs from a Nash equili-brium, it turns out that the self-confirmingequilibrium is also time-consistent. In thepaper presented, Sargent and Cho analyzethe dynamics of the system when the autho-rities use an approximating model with a‘constant gain’, recursive estimation algo-rithm, which discounts past observations.Under this new algorithm, there is anotherdeterministic component governing thedynamics of the system, called ‘escapedynamics.’ This component drives the eco-nomy away from the sub-optimal time-con-sistent equilibrium and occasionally leads ittowards the optimal time-inconsistent(Ramsey) outcome.

The intuition is the following. Whenthe system is in the self-confirming equili-brium, the government is in an experimen-tation trap. Although the econometriciandoes not realize the role of private expecta-tions in shifting the short-run Phillipscurve, his approximating model is able to

80

Thomas Sargent (Stanford University and Hoover Institution) and

In-Koo Cho (University of Illinois,Urbana Champaign)

Escaping Nash Inflation

Summary of the PaperAndrew Levine, Volker

Wieland and John C.Williams, from the Board ofGovernors of the FederalReserve System, jointly presented theirpaper on “The Performance of Forecast-Based Monetary Policy Rules under Modeluncertainty”. Their quantitative analysiscompared forecast-based monetary policyrules with outcome-based policy rules interms of output, inflation, and interest ratestabilization. Counterintuitively, their resultspromise only small benefits from the use offorecast-based rulesversus outcome basedrules.

The growing lite-rature on the quantita-tive evaluation ofmonetary policy rulesfor various macromodels includes therecent studies by Ball (1999), Batini andHaldane (1999), and Black et al. (1998), tomention a few. In a first approach, to com-pare alternative policy rules, Levine,Wieland and Williams (1999a) show thatsimple outcome-based policy rules performwell in stabilizing output, inflation, andnominal interest rate. The present analysisexpands their previous work by focussing

on the performance ofinstrumental forecast-basedpolicy rules in four macro-economic models that differalong the dimensions of

size, output and inflation dynamics, as wellas lag-structure.The models under conside-ration are: the Fuhrer-Moore (FM) model(1995), the MSR model by Orphanides andWieland (1998), Taylor’s multicountrymodel (TAYMCM), and the Federal Re-serve Board (FRB) staff model.The compa-rison of the results with the benchmarkvalues of the previously analyzed outcome-

based rules allow someconclusions to bedrawn:

1.) The ouput-,information-, and lag-encompassing featuresof forecast-based rules– which were assumedto generate an advanta-

ge over outcome-based rules – turn out tobe of minor quantitative importance.Thereis virtually no observable distinction in theoptimal policy frontier amongst the fore-cast-based and output-based rules in thesmall scale models (FM and MSR). In thelarge scale models (TAYMCM and FRB),the forecast-based policy rule seems to per-form only marginally better.

83

Greespan’s action may have been the outco-me of a process in which they were learningthe coefficient on inflation expectations,rather than the outcome of an escape dyna-mics.

In the second discussion, James Stockstressed four points. The first related to therole of the escape dynamics, which is trig-gered by the realization of a small numberof unfrequent and highly influential obser-vations. Stock noticed that, for the escapedynamics to prevail, it is important that theinformation set is restricted to a small num-ber of observations. One could also carrythe experiment of maximizing welfare withrespect to the number of observations to beincluded in the estimation, and the resultwould be a peculiarly low number, aroundfour or five.The second point was that, forSargent’s explanation to hold, the econo-metrician has to be rather mindless and cer-tainly not rational, since it doesn’t learn thetrue model. The third point related to theempirical evidence on the US inflation pro-cess. Estimating an auto-regressive processfor inflation over the period following thestabilisation, Stock found evidence for thelargest root being close to one.This findingis at odds with the evidence based on spec-tral analysis that Sargent presented, whichpointed at stationarity over that period.Thefinal point related to the theme of the con-ference. Stock noticed that, somehow para-doxically, uncertainty is beneficial inSargent’s model, because a lot of econome-tric uncertainty is needed to approach theoptimal Ramsey equilibrium outcome.

Several questions and comments wereraised from the floor. Bennett Mc Callumasked what type of dynamics emerges whenthe econometrician uses good econometrictechniques to estimate an expectationalPhillips curve.Tamim Bayoumi pointed outthat in Sargent’s explanation, as in the dyna-mic of scientific revolutions, some unlikelyevent would start a learning process. Animportant difference, however, is that inSargent’s model, the econometrician keepsestimating the same old model, even afterbeing at the Ramsey equilibrium. GuidoTabellini objected that the idea of havingsophisticated econometricians, who disre-gard available observations, is not veryappealing. David Vines asked whether therecent US experience, where for some rea-son policy-makers started reducing inflati-on despite their belief of being close to thenatural rate, could be seen as an example ofthe initial phase in the escape dynamics.

82

The Performance of Forecast-Based Monetary Policy Rules

under Model UncertaintyAndrew Levine,Volker Wieland

and John C.Williams(Federal Reserve Board)

past outcomes, it seemed that forecast-based and outcome-based rules wereredundant. He also noted that forecastsneed to take into account policy reactionfunctions, a feature that might render fore-cast-based rules trivial. The results are evi-dence for Glen Rudebusch (FederalReserve Bank of San Francisco) that fore-cast-based policy rules with a complicatedlag-structure in the state variables do notperform better than simple outcome-basedpolicy rules.

85

2.) If output stabilization is of minor impor-tance in the policy objective function, theperformance of forecast-based rules wor-sens. Instead, if minimizing inflation varia-bility is the sole policy objective, inflation-forecast-based rules prove to be efficient.

3.) Whilst outcome-based rules are gene-rally associated with stable and uniquerational expectations equilbria, forecast-based policy rules, with forecast horizonslonger than two years, often seem to gene-rate multiple rational expectations equili-bria.

4.) Forecast-based rules turn out to be rea-sonably robust to model uncertainty withthe limitations that forecasts are consistentwith the model and the rule itself , thepolicy rule does not require a forecast hori-zon beyond four quarters, and the inflationmeasure is a moving average over four quar-ters. The robustness of forecast-basedpolicy rules to model uncertainty disap-pears when the policy rules respond toexpectations in inflation more than a year inthe future, which is a common feature ofmany rules proposed in the literature.

Summary of the DiscussionCharles Bean (London School of

Economics) doubted the policy relevance ofthe multiplicity outcomes in forecast-basedrules. Instead, he suggested distinguishingbetween expectations of private agents andcentral banks. If a central bank communica-tes publicly the way it forms expectations,and private agents follow the central bank’sreasoning, they will settle at a unique equi-

libria and the multiplicity phenomena willdisappear.

Stefan Gerlach (Bank for InternationalSettlements) noted that the small benefitsof forecast-based rules over outcome-basedrules are not surprising. Many macroecono-mic variables follow low-order AR-proces-ses, which are also reasonablly good forecastrules. It is, however, not warranted to dis-miss the forecast-based rules since theirinformation-encompassing feature maybecome important should the low-orderAR-structure of macroeconomic variablesdisappear.

David Vines (University of Oxford) com-mented on the advantage of forward con-trol features in policy rules.The experienceof current account balancing, and the exi-stence of the J-curve has taught that feedb-ack rules are misleading when a complica-ted lag-structure initially inverts the res-ponse of the target variables. The multipleequilibria problem led Roger Farmer(European University Institute) to proposethat rational expectation monetary modelsbe supplemented with mechanisms thatdefine the way agents form their beliefs.These mechanisms work as selection devi-ces to pick an equilibria out of a multiplici-ty. Whilst Roger Farmer assumed a multi-plicity of equilibria, Bennett McCallum(Carnegie-Mellon University) raised thequestion whether the indeterminacy regi-ons are characterized by no solution, ormultiple solutions. Stephen Cechetti (OhioState University) asked for a clarification ofthe definitions. Since forecasts depend on

84

Summary of the PaperIn this paper,Athanasios

Orphanides analyzes Taylor-type policy rules in which the central bankresponds to inflation and the level of eco-nomic activity. In contrast to earlier studieswhich suggest that these rules provide aflexibility that yields substantial stabilizati-on benefits, the paper aims at showing thatthe macroeconomic policy failure in the1970s is due to an inappropriately activistmonetary policy, much like a Taylor rulewould have suggested at that time. Theapparent improvement in economic per-formance that activist monetary policyrules suggest over alternative policies,which completely ignore short-run stabi-lization objectives, can be attributed tounrealistic informational assumptions

regarding the knowledge policymakers canreasonably have about the state of the eco-nomy at the time when policy decisions aremade. Relying on counterfactual simulati-ons, the paper particularly points to mis-perceptions of the economy's productive

capacity by policymakers asthe underlying cause of the1970s inflation.

As a first step in his analysis,Orphanides theoretically considers theproblem of imperfect information by allo-wing for noise in the observation of thetrue rate of inflation and the true outputgap. This setup, following the Taylor rule,leads to undesirable interest rate move-ments that could adversely influencemacroeconomic performance. The authorthen re-establishes the stabilization promi-se of following activist policies by perfor-ming counterfactual simulations under theunrealistic assumption of perfect informa-tion. The results have the well-knownimplication that the inflation of the 1970s

would have been avoided if activist ruleshad been followed. However, by construc-ting a database with data available to U.S.policymakers in real time from 1965 to1993, it becomes evident that real-timeestimates of potential output severely over-

Athanasios Orphanides (Federal Reserve Board)

The Quest for Prosperity without Inflation

stated the economy’s capacity relative tothe recent estimates in the sample.Realistic policy rule simulations, which arebased on these real-time estimates, showthat activist policies not only would haveproduced the inflation of the 1970s, butwould have greatly inhibited the disinflati-on of the 1980s as well. Thus, the paperprovides evidence that 1970s inflation wasaccelerated not because policy must havedeviated from prescriptions suggested bythe Taylor rule, but because policy musthave actually followed a strategy indistin-guishable from the Taylor rule.

Orphanides concludes that apparentdifferences in the framework governingmonetary policy decisions during the 1970s,as compared to the more recent past, havebeen greatly exaggerated. To avoid policydisasters in the future, prudent policies likeinflation targeting, or natural growth targe-ting that ignore short-run stabilization con-cerns altogether, should be applied.

Summary of the DiscussionAt the beginning of his discussion,

Jordi Gali summarized the main findings ofthe paper. First, a simple Taylor rule basedon real-time estimates of inflation and out-put is able to characterize the behavior ofthe federal funds rate in the late 1960s andthe 1970s. Second, a persistent negativebias in output gap estimates can explain ahistorical interest rate level which was per-sistently below the level corresponding toprice stability. Third, a poor assessment ofmacroeconomic conditions by policyma-kers is responsible for over-expansionary

monetary policy in the past. The generallesson of these findings is that, because ofsevere data limitations, simple monetarypolicy rules do work well in theory but arenot appropriate for the conduct of mone-tary policy in practice. As a consequence,policy should rely on variables which couldbe measured relatively exactly. Thoughwelcoming the paper as a provocative studywhich will be widely cited, Gali addressedtwo specific issues with which he feelsunhappy, with regard to recent literature.The first issue refers to the notion of theoutput gap whose measurement is a keyfactor. Gali noticed that though empiricalresearchers have not yet resolved concep-tual problems measuring potential GDP,the paper is silent about this issue.While allavailable empirical concepts have the com-mon feature of modeling potential outputas a smooth time series, optimizing modelswith sticky prizes do not imply potentialGDP, defined as the natural output levelwhen prices are flexible, to be smooth.Referring to his own recent research, Galisuggested to couple micro foundations andempirical analyses by constructing a model-based measure of the output gap, using realmarginal costs. The performance of thisapproach in explaining inflation has beenshown to be very good. Regarding thesecond issue, Gali discussed the evolutionof monetary policy rules over time. Hestressed that monetary policy improved inthe 1970s and 1980s, which is supportedby recent empirical studies showing mone-tary response to inflationary pressures tobe varying over time. Since its focus is oninterest rate levels, the paper does not dis-

8786

Summary of the PaperThe objective of the

paper by Rudebusch is toshow how nominal incomerules for monetary policy conduct themsel-ves under uncertainty. The advent of theEuropean Monetary Union has again awa-kened interest in monetary policy strate-gies.The pillar of the ECB´s monetary stra-tegy shares some common features withnominal income rules. Any insights into theperformance of nominal income rules maytherefore be of interest to monetary policymakers inside the ECB.

Another source of the growing inte-rest in nominal income rules is the outstan-ding performance of the US economy. Thelong-lasting boom suggests, to some pundits,that a structural shift may have occurred.According to McCallum and Orphanides, itis the uncertainty surrounding the possibleoccurrence of a structural shift that consti-tutes the argument for conducting mone-tary policy within a framework of nominalincome targeting. The reason for this beingthat such a policy would not need to rely onestimations of the output gap, which arenotoriously disputable.

However, the advantages of nominalincome targeting must be weighed againstits disadvantages. Ignoring nominal income

rules when dealing with thedifferent time lags of mone-tary policy actions on inflati-on and output must be seen

as the main disadvantage. As Ball andSvensson have shown in a simple backward-looking aggregate demand and supplymacro-model, the performance of a nomi-nal income rule is somewhat poor, that is,the conduct of monetary policy followingsuch a rule leads to destabilization.McCallum criticized the results of Ball andSvensson on the grounds that their inflationspecification was backward-looking. In asetting with forward-looking specifications,McCallum finds nominal income targetingto be output stabilizing.

The controversy between McCallumand Ball/Svensson critically hinges on thequestion of model and parameter uncer-tainty. Rudebusch’s paper attempts to shedsome light on how nominal income rulesperform under uncertainty. In his simulati-ons he allows for uncertainty with regard tothe real output gap and the appropriatemodel. His different models are nested wit-hin a New Keynesian setting which allowsfor both forward and backward-lookinginflation specifications.

Rudebusch compares the performanceof three different monetary policy rules:

89

cuss this topic. This can be regarded as ashortcoming.

In the second discussion, Paul DeGrauwe characterized the paper as a fasci-nating study which shows how incompleteinformation can complicate monetarypolicy or even might lead to large policyerrors. After reconsidering the basic stra-tegy of the paper as simulating monetaryactivism, with and without perfect infor-mation, De Grauwe criticized that inflationand the output gap are the only two sour-ces of noise to which an activist strategy,with imperfect information, is subjected.This, however, would imply that the remai-ning data are perfectly known, which doesnot appropriately describe macroeconomicreality. An important example for othersources of noise is the inflation equationincluding cost push factors unobservable topolicymakers. Since the paper does notaccount for this kind of uncertainty, thereported results are biased against Taylor-type rules. De Grauwe further criticizedthat the simulation results rely on mindlessactivism, defined by him as mechanicallyapplying the Taylor rule over time, evenafter realizing large estimation and policyerrors. As an alternative, he suggested loo-king at additional indicators in order tosimulate intelligent activism, which mightlead to very different outcomes. DeGrauwe therefore concludes that the paperhas gone too far in criticizing activistmonetary policy.

Referring to his own conferencepaper, Lars Svensson questioned what hap-

pened to certainty equivalence. He arguedthat a comparison which relies on expostdata is not fair. He emphasized that poormeasurement of potential output can beeither explained by a wrong concept, or canbe biased because of real data. A late revisi-on could be interpreted as a signal for awrong concept. Vitor Gaspar pointed tothe significant output gap in the secondperiod of the analysis, which supports themain results of the paper. In his reply toJordi Gali, Athanasios Orphanides welco-mes more research on estimating the out-put gap. He also noted that the Taylor ruleis a sensible concept which does not descri-be mindless activism. Bennett McCallumstated that he has undertaken similar stu-dies using the same approach.A central fin-ding was the importance of measurementrevisions. John Taylor disagreed with thispoint of view.

88

Glen Rudebusch (Federal Reserve Board

of San Francisco)

Assessing Nominal Income Rules for Monetary Policy with

Model and Data Uncertainty

two nominal income rules based onMcCallum and Orphanides and one simpleTaylor rule are the benchmarks. To derivethe optimal policy rule, he uses a standardobjective function which minimizes variati-ons in inflation and the output gap.

To achieve robust results regarding theperformance of each policy rule, Rude-busch calculates the objective functionsover a range of different model specificati-ons and data uncertainties. The assumedmodels thereby differ in the way inflationexpectations are incorporated. Each modelspecification is weighted with its appropria-te probability. Data uncertainty comes intoplay through uncertainty regarding the out-put gap.The result of his simulation shows arather poor performance by the nominalincome rules of Orphanides and McCallumas compared to the simple Taylor rule.Thisresult holds for uncertainty concerningmodel specification, as well as data uncer-tainty.

Summary of the DiscussionBennett McCallum voiced strong

reservations in his discussion about the con-clusions drawn in Rudebusch’s paper. First,he stated that the results of his paper aremore general than stated by Rudebusch.Instead of just one inflation specification,his results were derived for 7 differentprice adjustment specifications. Further-more, he disagreed with Rudebusch’s claimthat nominal income rules show a disa-strous performance. Should this attributebe given to the two proposed nominal inco-me rules, it should be assigned to actual

monetary policy in the United States aswell. The reason for this is that actual dataperform worse than both nominal incomerules when applied to Rudebusch’s loss fun-ction. Finally, McCallum criticized therange of parameters Rudebusch had chosento specify price setting. He points to a paperby Gali and Gertler which suggested thatRudebusch’s set of admissible parameterswere too narrow. Further model specificati-ons, like forward-looking demand or anopen-economy setting, should have beenchosen as well.

Henrik Jensen, the second discussant,appreciated Rudebusch’s paper as an “excel-lent, clear, and interesting paper”. Jensensees particular merits in the robustness ana-lysis since it shows that an optimal Taylor-rule dominates both proposed nominalincome rules under output-gap and modeluncertainty. Nevertheless, he sees therobustness analysis as being incomplete andtoo favorable towards the Taylor-rule. Theintuition behind Rudebusch’s results can beexplained by the fact that the Taylor-rulecontains the output-gap as a welfare rele-vant variable, whereas both nominal inco-me rules contain “almost” no welfare-rele-vant variable. Nonetheless, both nominalincome rules perform well as long as for-ward-looking behavior in price-setting rea-ches a certain level.This lead Jensen to pro-pose a “flexible” nominal income rule.

During the open discussion after-wards, Orphanides stressed the importanceof learning, whether to incorporate the out-put level or the growth rate, when assessing

the performance of policy rules.Vines men-tioned that nominal income rules do nothave a good dynamic performance. Toimprove the robustness analysis, perfor-mance under the occurrence of supply andtrade shocks should be investigated as well.Svennson mentioned that two interpretati-ons of nominal income targeting exist. Hetherefore asked the proponents of nominalincome rules which of these concepts wasactually meant. Taylor highlighted theimportance of potential output uncertaintywhether monetary policy is conducted on aday-to-day basis, or rule-based. Levin que-stioned the justification of the restriction ofparameter equality which nominal incomerules impose on the output growth andinflation coefficient. Nelson criticized theNew-Keynesian setting in Rudebusch’spaper. He claimed that in a setting withcapacity being time-varying, nominal inco-me rules would outperform a simpleTaylor-rule.

9190

Summary of the DiscussionAlex Cukierman argued that this was

an interesting as well as thought-provokingpaper. The central bank not only faces theusual uncertainty, but also uncertaintyabout the effectiveness of its policy. Crucialis the question of how much the centralbank is able to learn. A central bank mustexperiment. This leads to the notion thatkeeping the actions of the central bank sta-ble, allows it to have more sample points forthe central bank. However, because of stra-tegic interaction, this is not necessarily thecase. In periods of low policy effectiveness,the inflation bias will be small and thepolicy-maker has an interest in revealingthis. In periods of high policy effectiveness,the inflation bias will be high in which casethe policy-maker has an interest in keepingthis undisclosed. If the latter effect is stron-ger, the hiding mechanism, dominates. Themain result of this paper is, therefore, thatan additional mechanism apart from theBrainard conservatism principle, existswhich leads to fewer actions by the centralbank.

Summary of the Paper “Transparencyand reputation: Should the ECBpublish its inflation forecast?”, PetraGeraats.

Before presenting the model develo-ped in the paper, Petra Geraats introduceda framework to identify the differentdimensions of transparency that are rele-vant for monetary policy. Political transpa-rency refers to the openness about policyobjectives, economic transparency on thedisclosure of economic data, procedural

transparency on the policy decision-makingprocess, policy transparency on the com-munication of present and future decisions,while operational transparency touchesupon the implementation of policy andmarket interventions.

The paper addresses the issue of eco-nomic transparency, and more specificallythe advantages and disadvantages for a cen-tral bank should it release its own inflationforecasts to the public.To do so, Geraats haselaborated a two-period, Barro-Gordonmodel, in which the public interprets theinterest rate setting by the central bank as asignal of its commitment to control inflati-on. In the «transparency»-dominated set-ting, the first period interest rate is a lessnoisy signal than in a setting of «opaquen-ess». As a result, markets respond more tothe interest rate signal if the central bank istransparent. In the latter case, the incentivefor the central bank to invest in reputation,by targeting lower inflation in the first peri-od, is relatively smaller, and the realised firstperiod inflation is higher than in the regimeof transparency. The model also shows that,in addition to the reduction of the inflationbias, transparency increases the leeway forthe central bank to stabilise the economy.

Finally, as the result of the model relieson the ability of the central bank to revealwhat is the structure of shocks affecting theeconomy, the paper shows that this can bet-ter be achieved through the publication ofconditional, rather than unconditional,forecasts.

93

Summary of the Paper “Learning, uncer-tainty and Central Bank Activism in aneconomy with Strategic interactions”,Martin Ellison and Natacha Valla

In this paper the authors construct astylised model of monetary policy in whichthere are roles for uncertainty, learning, andstrategy. The model they construct is cha-racterised by an expectations-augmentedPhillips curve between inflation surprisesand output deviations from trend. Inflationis assumed to be completely under the con-trol of the central bank, and is the instru-ment of monetary policy.

Uncertainty in the model takes theform of a two state Markov-switching pro-cess for the monetary policy effectiveness.Monetary policy switches between high andlow policy effectiveness. The central bankhas an informational advantage in the formof receiving a signal of the output shock.This informational asymmetry creates abasis for stabilisation actions.

The actions of the central bank lead tooutcomes from which the central bankupdates its beliefs about the state of mone-tary policy effectiveness.The same learningprocess that takes place in central banks alsogoes on at private agencies. Since privateagents can deduce from the policy actionsthe output signal the central bank received,

learning is symmetric. Private agents andcentral banks have identical beliefs.

The authors then consider two equili-brium concepts. In the first one, the centralbank is a passive learner.A central bank thatis passively learning will optimally take intoaccount current uncertainty but does notmake any conscious attempt to influencesuch uncertainty in the future. It does notconsider that policy actions can be used toupdate beliefs.The second equilibrium con-cept considers the central bank to be anactive learner. The central bank takes intoaccount current uncertainty and is awarethat actions taken today generate informati-on which affects future uncertainty. Afterestimating the model for the G7 countries,the authors calibrate the model to performsimulations. The main result is that anactively learning central bank reacts less toan output signal than a passively learningcentral bank would.The reason for this resultis that the degree of central bank activismdirectly determines the degree of learning,in both central bank and private agencies,Due to strategic interaction between thecentral bank and private agents, the centralbank has an incentive to speed up or slowdown the learning of private agents. Whenmonetary policy is effective, the centralbank prefers slower learning to avoid agentsadjusting their inflation expectations.

92

Summary of the Presentations and Discussions during the

Short Papers Session

those aspects of transparency are someti-mes in conflict with each other. Onecorollary is that, in general, it will no lon-ger be possible to rank different monetarypolicy strategies and communication poli-cies in any meaningful way.The paper also discusses the persistent con-troversy over ECB transparency. It explainsit by the extraordinary communicationchallenges faced by the ECB, which mustcommunicate to fragmented audiences in amulti-cultural environment.

Summary of the DiscussionMarvin Goodfriend argues that this

paper is about the economics of conversati-on with various audiences. Central to theissue is the fact that the central bank has tocommunicate with different market partici-pants: financial markets, price-wage set-ters, households, foreign investors, etc.Thecentral bank cannot informationally discri-minate in the same sense a price setterwould price discriminate. Therefore, com-munication becomes a very difficult job.Sustaining a conversation is further hampe-red by the fact that the current economicsituation continuously changes. In addition,historically the public at large and the presshave come to know and understand more,which requires more from the central bank.He concludes that there is a role for thecentral bank in educating and teaching thepublic.

Charles Freedman elaborated furtheron the limits of transparency. He questio-ned which type of statements actuallyrevealed information and which ones mere-

ly added noise. He also posed the questionof what to publish: conditional versusunconditional forecasts. He referred to theBank of Canada experience in which themarket interpreted forecast conditioning asan intent on the part of the central bank.Problems may arise if the central bank thenhas to deviate from the forecasts when con-ditions change. He also pointed to the factthat different audiences have differentunderstandings.

David Vines argued that the paper wasnot only about conversation, but also aboutcentral bank action. He pointed towards theinescapable conflict between openness andclarity.The actions of the central bank couldnever be fully described by an algorithm.He pointed out that there will always be anunexplainable epsilon.

Summary of the Paper “Caution andConservatism in the making of Mon-etary Policy”, Philip Schellekens.

The paper explores the robustness ofthe key results of the credibility literature,from time inconsistency inflation bias todelegation, with multiplicative uncertaintyand generalised quadratic central bankobjectives. In the model, the supply curvetakes the traditional Lucas form and thetransmission process, from policy instru-ment, to realised inflation is given by a sto-chastic multiplicative factor. In addition, thequadratic loss function of the central bankdeparts from the one used in the credibilityliterature by admitting different weights formissing inflation or output targets and fortheir variability.This allows the aversion for

95

Summary of the DiscussionAlex Cukierman began his discussion

of the paper by summarising the key fea-tures of the model, such as: the public igno-res the true inflation target of the centralbank and knows that its policy will not betime-consistent. In this context of asymme-tric information, there is an incentive to betransparent so as to improve the ability tostabilise the economy. Cukierman askedabout the relevance of assuming that thepublic ignores the inflation target of thecentral bank.

The second discussant, Marvin Good-friend, broadened the debate by putting thepaper in the perspective of central bankpractice in the real world. He first questio-ned the relevance of models relying on out-put inflation trade-off for central banks,such as the ECB, with a lexicographic pre-ference for inflation. He then concentratedon the drawbacks of publishing forecasts.Tostart with, unconditional inflation forecastsby central banks who target inflation areconstant so that it would be meaningless topublish them on a regular basis.The uncer-tainty of the forecast may then confuse themarket. The market expectation of futureinflation, which should be as good as theone elaborated by central bank economists,is readily available in the yield curve.

In the general discussion, Adam Posencriticised Geraats’ model, and more gene-rally models of time inconsistency, on thebasis of empirical evidence that shows thecentral banks’ ability to reveal their prefe-rence to the public.

Summary of the Paper: “Which kindof transparency?”, Bernhard Winkler

In this paper, Bernhard Winkler deve-lops a language in which central bank com-munication can be meaningfully discussed.He develops a theory on the meaning of thedifferent uses of the word ‘transparency’.

The paper first surveys the literatureon transparency in monetary policy-mak-ing. This literature has given precise mea-ning to the term ‘transparency’ in the con-text of models, which do not analyse thecritical issue of how to best communicatemonetary policy. Winkler questions theassumptions of perfect information proces-sing by individuals, common knowledge ingame situations, and perfectly efficientmarkets.

In his paper, Winkler proposes todistinguish different aspects of transparen-cy.The first aspect, clarity, refers to the factthat information needs to be simplified,processed and interpreted, in order to beunderstood. The second aspect, honesty,requires that communication correspond tothe true intended meaning of the sender,and this is independent of whether it is clearor understood by the receiver. The thirdaspect of common understanding refers tothe need for the sender and receiver in themonetary policy game to share a commonperspective and have sufficient knowledgeof each other’s information.

Winkler further develops a transpa-rency triangle consisting of honesty, clarity,and information efficiency. He argues that

94

from the target, the uncertainty about theirfuture development is greater. Since thepersistence of inflation only affects thedynamics of the economy, optimal policyreduces the amount of uncertainty aboutfuture inflation by acting more aggressivelyto push inflation closer to target.

A further result, so the paper, is thatparameter uncertainty can lead to smoo-ther paths of the interest rate than undercertainty equivalence. This mitigates theneed to explicitly enter a smoothing objec-tive into the central bank’s loss function.

Summary of the DiscussionCarl Walsh compared the paper by

Soderstrom to the one by Schellekens. Hediscussed what one can learn from thosepapers regarding uncertainty, learning, andstrategy. He stated that Soderstom had awider set of parameter uncertainty thanSchellekens, where only one parameter wasuncertain. He argued that the exogeneityassumption of the uncertainty might not becompletely revealing. In the real world,much of the parameter relates instead tothe lag of the transmission mechanism.More interestingly, the comparison of theU.S. disinflation in the early 1980’s, and theongoing 1990’s disinflation suggests thatthis lag is, itself, an endogenous response bythe private sector to the monetary policy-makers. One may then regret that themodels of Schellekens do not tackle thelearning mechanism.

He argued that monetary policy itselfplays a role in creating parameter uncer-

tainty.There is nothing to learn about lear-ning itself in the models of Soderstrom andSchellekens. Because of the backward-loo-king features of the Soderstrom modelthere is no role for strategic interaction.Ideally, expectations should be incorpora-ted in the model. He further argued thatthe paper did show that uncertainty inmodel dynamics removes the presumptionthat caution is optimal. However, the realnet effect of overall parameter uncertaintymight not be that large.

97

missing targets to be separated from theone relating to macro-economic instability.

First, Schellekens shows that multipli-cative uncertainty introduces an incentivefor the central bank to be cautious and thatdelegation to a conservative central bankerimproves general welfare. Second, he offersa new way to approach conservatism in thedelegation of monetary policy. He finds thata more flexible definition of central bankpreferences helps to release the trade-offbetween flexibility and credibility. In otherwords, the appointment of a conservativecentral banker does not imply relinquishingthe ability to stabilise output.

Summary of the Paper “MonetaryPolicy with Uncertain Parameters”,Ulf Soderstrom

Soderstrom develops a dynamic macro-economic model in which there is uncer-tainty about structural parameters. Heshows that this uncertainty does not neces-sarily lead to more cautious monetarypolicy, contrary to the current wisdom ofthe Brainard conservatism principle. In par-ticular, when there is uncertainty about thepersistence of inflation, it is optimal for thecentral bank to respond more aggressivelyto shocks than under certainty equivalence.This is so because this way the central bankreduces uncertainty about the future deve-lopment of inflation. Uncertainty aboutother parameters acts to dampen the policyresponse.

For his analysis, Soderstrom uses thedynamic aggregate-supply, aggregate-demand

framework developed by Lars Svensson.The model consists of two equations rela-ting the output gap and the inflation rate toeach other, and to a monetary policy instru-ment. The central bank has the traditionalquadratic objective function. Monetarypolicy is assumed to affect the output gapwith a lag of one period which, in turn,affects inflation in the subsequent period.Soderstrom modifies the Svennsson frame-work by assuming multiplicative uncertain-ty. The parameters of the model are assu-med to be time varying. They are randomvariables with time invariant means. In thismodel, certainty equivalence ceases to hold,and the variances of the state variables affect the optimal policy rule.

Soderstrom solves the model numeri-cally. He then analyses how the resultingpath of the monetary policy instrumentdepends on the degree of uncertainty of themodel. To analyse this, he considers firstuncertainty about the model parametersseparately. In a second stage he combinesuncertainty of more than one parameter. Inthe case of impact uncertainty, the uncer-tainty about the parameters in the transmis-sion mechanism, the Brainard conservatismprinciple still holds.This is also the case foruncertainty about the persistence of output.

However, uncertainty about the persi-stence of inflation, causes the central bank toact more aggressively. One reason for thisresult is that under multiplicative uncer-tainty, the variance of the model variablesincreases with the distance from target.When inflation and output are further away

96